portalbasic searching → images.htm
 You'r deep inside searchlores 
This is a windrose
Images & Pictures Searching (and more :-)

         First published @ www.searchlores.org in Autumn 2003   ~   Updated by fravia+, March 2009, version 13.5

Images searching


[Introduction]
[Bypassing copyrights hassles]
[Names, names everywhere]
[Where to search for images]
[Formats & downloads]

red [Images searching: various tools and repositories]     
red [Images searching: main engines]      

[Death of the "edu" trick]
[Altosax's "not-image" trick]
[Finding images: a discussion]
[Images' databases]

red [Images' essays]     

[Images' semantics]    [Stalking through images]

    
images searching

Labs and workshops
Searching images without even knowing their names
[The Caravaggio example]
[faf_task.htm] ~ [ra_task.htm]
[loki's emperor's clothes]
[Jorge Luis Borges' tomb]
Finding and understanding images without even knowing their names and purposes
"The chick on the river"

Searching images on the wide web


Images searching: an introduction
Also: how to (try to) avoid copyright hassles when publishing
(by fravia+)

The Web hosts billions (milliards) of images. How many is anybody's guess. Most seekers calculate that the total number of images on the Web has exceed 20 billions (milliards) during 2004. Should have been around 40 billions (milliards) in 2007 and 80 billions right now (March 2009). These numbers must be taken cum grano salis, they are however on the conservative side (Flickr alone, launched in 2004, has atm 4 billions images).
Only about 10% of these images are indexed by the images searching engines. You may spend a long time looking for the remaining 90%, and you are probably deemed to fail, unless you learn some of the tricks explained inter alia on this very page.

I firmly believe that every picture, film, photo, sketch, drawing, logo, mosaic... in short every image that humanity has ever concocted - from the noblest to the most perverse sense of this verb - is already on the Web, somewhere, (so far as it has been published in the last 50 years on any media that could have been digitized, duh).
But do not take my word for it, try it out! Upload somewhere an image scanned by yourself... let the web simmer for some months... shake well, and then search that very image again using the techniques described below: chances are you will find several copies of your image elsewhere.
Should -exceptionally- a given image still being amiss (after a thorough search), do not worry: it will eventually land on the web as well, sooner than you may expect :-)

Put your hear to the searchscape and be attentive... see? Whole pinacothecae (picture galleries) are going on line in this very moment: the complete works of some barely known asian photograph, the Turner collection at Tate's, a complete scanned comics collcection of the fifties, the whole images archive of -say- the Jornal do Algarve... you name it, your eyes will have it.

Everything is there... but where? Where is THE image that you need now, and you mean now, on the huge web? Will you be able to find it effectively and quickly? Well, you will enjoy hearing that the answer is: "Most probably".

There are various problems, though: First of all the amount of trash currently on the web is mind-boggling. Most of it, how you may have already realized, is due to an extremely vulgar paroxism of 'commercialisation'.
Search for some specific artist's pictures, using the main search engines, or their specific image searching facilities collected on this page, and you will most probably be smothered in a squalid net of "reproduction" sellers.

Their purpose is of course NOT to reproduce on their sites in the best possible way the pictures you are looking for, their purpose is to sell you some cheap reproduction and/or to have you pay for the 'right' to look at the picture you are interested in.

A small digression: images as they should be

There are some exceptions, though: for instance "googleearth" (of all places) offers for free some incredibly detailed reproductions of (few: only 14) paintings from the Prado Museum. These images have a resolution of 14,000 megapixels (some 1,400 times greater than a picture taken on a standard -good- 10 megapixel camera). You have to install "googleearth" first, then go there to "layers", then geographic, then prado. You can of course then use Gimp to grab the details you are interested in. Worth seeing in order to understand how ridiculous are the attempts of those that want to "sell" reproductions and how beautiful the web could be, if only the morons that manage our museums as if they were private property would for once do their job and spread culture instead.
These images were sewn together digitally from more than 8,000 high-resolution photographs of sections of the paintings.

Anyway in most cases your first hurdle will be to diminish or if possible completely eliminate the commercial noise in order to get a cleaner signal of your target. Even the simplest tricks (for instance just adding -".com" to your searchstring) will go a long way to diminish the noise. You may also want to study more complex approaches in your quest towards seekers' perfection.

Also the huge amount of commercial PORN on the web can compel us to walk some tricky paths in order not to drown inside a quicksand of unwanted images (in fact you'll have to fend off tons of unwanted commercial crap even when looking purposely for a specific porn image). You'll soon realize that the 'pornography' problem of the net is NOT AT ALL due to the real porn images collections per se, but rather a consequence of a zillion of aggressive commercial sites triyng to scrap some bucks - and polluting the whole web - using the most awful, vulgar and trite zombies' bites.

When searching for images you will often have to deal with (read "enter") huge databases of images which have been puposedly 'hidden', and that you (theoretically) should only access once given permission by -say- a greedy commercial entity or some malbehaving public museum, slowly sinking towards its own privatisation sunset.
In such cases some simple and well-known password bypassing tricks can do wonders.

Finally, note that images searching might be based on
  1. images content (e.g.: Exalead, Pictitup and googleimages)
  2. images color (e.g.: Pictitup, picsearch, multicolr search lab and etsy)
  3. images similarity (e.g.: tineye)
    "TinEye is a reverse image search engine indexing more than 1 billion (milliard) images. You can submit your own image to TinEye to find out where it came from, how it is being used, if modified versions of the image exist, or to find higher resolution versions. TinEye is the first image search engine on the web to use image identification technology rather than keywords, metadata or watermarks"



Copyright hassles

Granted: the following could possibly provoke some raising of eyebrows among the most copyright conscious corrupt politicians of ours. So you should better gain some basic knowledge of anonymity lore and password lore, just in case.

Indeed -and specifically in case of images- a further (albeit silly) problem is the growing "copyright histeria" pervading our american masters and their local political lackeys. Suffice to say that you still can (at least in the European Union) fetch, use, stomp, burn, modify, alterate ANY image published on Internet provided that you respect some points: non profit use in the context of a creative work, and of course giving due credit.

You better harry up, though, all legal screws are being tightened in order to nail inside its coffin any spark of knowledge and creativity that should still happen to survive the "trendy" commercial perversion we have to endure.

From a theoretical point of view, when citing Images and Pictures that you wish to use in your own papers or essays, you should cite the following elements:
  1. Artist's Name, if known
  2. Copyright's holder, if known
  3. Title of the image, if known (if not, use a description)
  4. Institution where held, if known.
  5. Title of article or book (if applicable)
  6. Author of article or book (if applicable)
  7. Title and Date of magazine (if applicable)
  8. Database name (if applicable)
  9. Date of access if online or publication if originally from print material URL (if applicable)
Clearly some of these requisites, and especially the second, seem (and actually have been) devised in order to stiff creativity.
I know of a very siple solution to such images' copyright hassles: the following (this is all just theoretical academy, he, you will do this only if compatible with your local petty and heavily corporate lobbied anti-creativity laws).

Let's imagine (imagine) that you badly need an image that might have a copyright owner somewhere (or might not... let's just say you tried to contact this "owner" but failed).
In such a case let's imagine (imagine) that you:
  1. copy the image on your harddisk
  2. pass it with GIMP from jpg or png format to an uncompressed format and then back to png with a different compression ratio
  3. upload and publish it on any free page provider or blog somewhere on the web under a fake identity made through your cousin or devised on the fly in an internet cafe
  4. state clearly on this new site/blog that the image is free of any copyright and can be used by whomever for whatever purpose
  5. cite "in good faith" this site/person when you publish it on your essay/book
Quod erat demonstrandum.

Names, names everywhere

So, once again, every image is "still and nevertheless" there... but where? Where is THE image that you need now, and you mean now, on the huge web?
Alas, the web -as we well know- is a web of WORDS. Hence the way you prepare your "search arrows", the way you formulate your search strings or your search scripts is of tantamount imporance in order to find effectively (or even at all) the images you are seeking.
Word and images are an INTEGRATED WHOLE! Hence inter alia the importance of the file formats discussed below, which are of course a part of the URL (web-address) of your target.

As you may imagine, the problem onto a "web of words" is that most images gathering scripts and most search engines, in order to identify image files within the web, just examine the very "words" that happen to be in the title or in the proximity of a given image. (Note however that google's &imgtype=face parameter, explained below, is a first step towards an interesting different approach).
The spiders look first at standard file extensions such as "jpg.", and just integrate, additionally, the image's file name and any eventual information they may be able to gather from the path. Bingo! That's all, thankyou very much.
Unfortunately, file names and path names are often very cryptic, frequently abbreviated, and they - more often than not - hardly describe the visual content of the image.
Hence if, say, a white rabbit has been named "snowflake", you may find that your results will be littered with a bunch of pictures of someone's favourite pet.

Even worse than cryptic names are -for seekers- those 'generic' image names, for instance image1.jpg, home.jpg and logo.gif: field examples of web-triteness. In our society thou shalt never underestimate thyne neighbour induced propensity for utter banality.

Where to search for images

An image or picture query is usually started using the general images searching engines facilities, offered by main search engines like Google or Altavista. It is also useful to perform searches on specialised Images searching tools, that index only images or multimedia.
Images public repositores la flickr might offer also some results, ditto for file repositories la rapidshare.
In addition, there are also images metasearch engines, which pass on search requests to more than one search engine and then bring back the results. (For such engines you should head the same caveats valable for all compound and metasearch engines: slightly better recall, but general slowness and lack of precision).
Searching through blogs, and of course on usenet can also give excellent images results.

The usual combing and local searching techniques used for general searching of course apply to images searching as well.

But before even beginning, you might wish to get acquinted with the different and most common image formats.


Common Images formats: PNG, JPG, GIF


Some info about the MAIN formats used on the web for images: JPG, GIF and PNG can be useful for images seekers.
While GIF and PNG are great for computer generated images with limited palettes, PNG and JPG are far better for photographs, because give better quality images for the same file size. PNG should always be used, whenever possible, because of its patent-free compression algorithm.

GIF (Graphics Interchange Format) is a data stream-oriented file format used to define the transmission protocol of LZW-encoded bitmap data. GIF images may be up to eight bits (256 colors) in depth and are always compressed. Despite the fact that GIF supports only 8-bits worth of colors, and the multimedia extensions introduced in the 89a release have not been widely utilized, GIF still remains a popular choice for storing lower resolution image data.
Specifications: ftp://ftp.ncsa.uiuc.edu:/misc/file.formats/graphics.formats/gif87a.doc; ftp://ftp.ncsa.uiuc.edu:/misc/file.formats/graphics.formats/gif89a.doc; http://www.w3.org/Graphics/GIF/spec-gif87.txt.

JPG (in dos systems) and JPEG (mostly in non-dos systems)
(JPEG-JFIF "Joint Photographic Experts Group" - "Jpeg File Interchange Format")
is optimized for photographs and similar continuous tone images that contain many, many colors. GIF compression is unkind to such images. JPG works by analyzing images and discarding kinds of information that the eye is least likely to notice. The JPEG format stores information with 8-bit per color - red, green and blue for a total of 24-bits.
The degree of compression is adjustable. At small compression levels of photographic images, it is very difficult for the eye to discern any difference from the original, even at extreme magnification. Compression factors of more than 20 are often quite acceptable. Better (free) graphics programs, such as Gimp or Paint Shop Pro, allow you to view the image quality and file size as a function of compression level, so that you can conveniently choose the balance between quality and file size. Note that JPG/JPEG files suffer generational degradation when repeatedly edited and saved.
Specifications: ftp://ftp.uu.net/graphics/jpeg/jfif.ps.gz; ftp://ftp.uu.net/graphics/jpeg/jpeg.documents.gz

Currently, GIF and JPG are still the formats used for most web images.
However, PNG (Portable Network Graphics) does everything GIF does, and better, being not limited to 256 colors, so many expect PNG, created as the free and open-source successor to the GIF file format, to replace GIF.
You should always use PNG instead of JPG for your own images and pictures because of the following advantages: Lossless compression, 48-bit, true-color depth, varying levels of transparency, gamma correction, better compression (10-30 percent smaller files than GIF), searchable content, patent-free compression algorithm.
The PNG file format supports true color (16 million colors). PNG excels when the image has large areas of uniform color.

The advantage of PNG on JPG/JPEG is that saving, restoring and re-saving a PNG image will not degrade its quality. PNG does not support animation like GIF does, though.
The PNG format provides a portable, legally unencumbered, well-compressed, well-specified standard for lossless bitmapped image files. Although the initial motivation for developing PNG was to replace GIF, the design provides also some useful new features not available in GIF, like opacity (PNG allows the control of the degree of transparency).
Specifications: http://www.w3.org/TR/REC-png-multi.html, second edition: http://www.w3.org/TR/PNG/

Another, for seekers less important format, used mostly for archives and desktop publishing, is the TIFF format (Tagged Information File Format, aka Tagged Image File Format aka... See http://home.earthlink.net/~ritter/tiff/).
It is a flexible image format that supports either 16-bit per color - red, green and blue for a total of 48-bits - or 8-bit per color - red, green and blue for a total of 24-bits - and uses a filename extension of TIFF or TIF.

Images searching: various tools and repositories


The Web Gallery of Art
The Web Gallery of Art is intended to be a free resource of art history primarily for students and teachers.
FUNDAMENTAL tool for finding paintings.


(The form below might not work because of google stupid malware services. If that's the case, just click on search once you are at the web gallery and query again)


  Author: Text: Time: Form:



PICASA
Picasaweb, google's picasa collections:
http://picasaweb.google.com/lh/searchbrowse?q=snowflakes&psc=G&filter=1#0+1



flickr
Doesn't show the height*width of the images in the SERPs.

Find a photo of...

 Perhaps you'd like to browse everyone's photos?




Ditto
Very commercial, unfortunately, but Ditto still carries a good collection of pictures.
Formerly known under the name Arriba.
Doesn't show the height*width of the images in the SERPs.

Search: 



Picsearch
"... all offensive material is filtered out by our advanced filtering systems..."
Shows the height*width of the images in the SERPs.

http://www.picsearch.com/

 Basic Search - Search Help -
 

 Only Images
 Only Animations
 Images and Animations
   Only Color
 Only Black&White
 Color and Black&White
Pictures' size must be



The Amazing Picture Machine

[http://www.ncrtec.org/picture.htm] (The Amazing Picture Machine)


Webseek (Columbia uni)

"...a new effort to catalog the visual information on the World Wide Web. WebSEEk has so far catalogued over 650,000 image and 10,000 videos from many sites on the WWW..."

You can browse subjects as well

Image/Video Topic (single word)
   
   
all
   
videos
   
color photos
   
gray images
   
graphics



Nicolas Pioch's Webmuseum (paintings)

Some companies may be trying to get a monopolistic grab on arts and culture, developing a pay-per-view logic, shipping out CD-ROMs while trying to patent stuff which belongs to each of us: a part of our human civilization and history.
This exhibit is not trying to compete in any way with books or specialized CD-ROMs. Such an Internet exhibit will neither reach the quality of paper reproduction and professional critic, nor will it be as easily available as a local CD-ROM, given the transfer time on the Internet.
No support, no funding, no manpower: the WebMuseum is a collaborative work of its visitors contributing to expand and improve the WebMuseum.


http://www.ibiblio.org/louvre/paint/


UNESCO

The collection -- covering a wide range of subjects related to the Organization's fields of competence: education, science, culture and communication -- was started in 1946 when UNESCO was founded and currently contains over 10,000 digitalized images. More will become available as the rest of the collection is digitalized and further developed so as to ensure a more equitable balance of subjects and countries.

You can search per Region, Country, Place, Theme, Subject, Keyword, Year or Photographer, and you can also display Thumbnails first.

Images repositories


http://www.fapomatic.com/, http://www.imghost.com/, http://www.glowfoto.com/, http://www.imageshack.us/, http://www.imgspot.com/, http://www.mytempdir.com/, http://www.bestupload.com/, http://www.netpix.org/, http://www.jotapeges.com/, http://www.rapidshare.com/, http://www.filesupload.com/, http://www.updownloadserver.de/, http://www.dropload.com/, http://www.sendthisfile.com/, http://www.fireupload.com/, http://www.yousendit.com/, http://www.youshareit.com/, http://www.glintfiles.net/, http://www.paintedover.com/, http://www.2and2.com/, http://www.imagehosting.com/, http://www.xs.com/, http://www.imagehigh.com/, http://www.imagevenue.com/, http://www.shareitagain.com/, http://www.ultrashare.net/, http://www.sendmefile.com/, http://www.perushare.com/, http://www.megaupload.com/, http://www.imageranch.com/, http://www.photobucket.com/





Special collections (just to show how deep is the web)


Henry Gray: Anatomy of the Human Body (Illustrations, 1918)

http://www.bartleby.com/107/indexillus.html

http://www.ars-grin.gov/npgs/images/cor/pwc/
(Pomological watercolors)
For instance winfield1

Three dimensional images - 3D images searching


[Three dimensional Objects Search] (does not work anymore)

http://web.archive.org/web/20030622052315/http://www.deepfx.com/meshnose/ (search engine still works from webarchive)

Usenet images retrieval
Pictureagent [pa27.zip] : 2716159 bytes, version 2.7
Fundamental tool for usenet image retrieval, 'protected' with one of the stupidest protection you can find around: the program carries a long hardcoded list of invalid names (probably people that did spread their registration codes on usenet) and cracker names like "hans die wurst", "donald ibarra", "escom/core", "iceman [ucf]" and so on, followed by this silly codesnippet:
:41596D 8B442430           mov eax, dword ptr [esp+30]
:415971 3BF8               cmp edi, eax
:415973 757A               jne 4159EF  ;bad_guy_exit
:415975 85FF               test edi, edi
:415977 7476               je 4159EF   ;bad_guy_exit
:415979 83F801             cmp eax, 1
:41597C 7C71               jl 4159EF   ;bad_guy_exit
:41597E 8D4C2410           lea ecx, dword ptr [esp+10]
:415982 C644242400         mov [esp+24], 0
and, ahem, you actually might want to 'nop' the 4159ef 'bad guy' locations :-)
When will programmers learn [some better tricks] to protect their software?

Images searching: main engines

Google  Altavista  Ixquick  Baidu  Fast  Excite  Lycos  Yahoo  Fotosearch 
Freenet  Latin Guya  Webshots 


Google Image searching

June 2007: 77200 results.

Google's ADVANCED Images search

You can use the advanced image search page to specify a series of parameters
Alternatively you can add on your own to the URL the relevant stringsnippets listed below: per hand or per bot.
Google is now the ideal images search engine to search for faces, duh.

        Advanced Search    Preferences    Search Tips

  Search images  Search the Web    "Safe search" and/or "Mature content filter" is Off


Google shows thumbnails (small versions of the original images) with the filename, size of the image in pixels, and file size. Only 20 results are shown on each page.
Note that if Google finds more than one image from the same web site it only shows the first one. Click on the link [More (results) from...] to see the remaining images.

Altavista's Image search

June 2007: 146,582 results.

[http://www.altavista.com/image/default]
The following query string has already as a target the painter "bilibin".
To create your own "quick" scripts, just substitute per hand or per bot "bilibin" with whatever imagequery you are interested in.
http://www.altavista.com/image/results?pg=q&stype=simage&imgset=2&q=bilibin

Here the different parameters (the ideal images search engine to search for backgrounds and wallpapers, duh):
  • Kind: Only photos &mik=photo  only graphics &mik=graphic  only buttons &mish=all  everything: &mik=photo&mik=graphic&mish=all 

  • Colors: only color &mip=color  only black and white &mip=bw 

  • Sources: web &mis=theweb  news &mis=news  movies &mis=ymovies  autos &mis=yautos 

  • Dimensions: small &miwxh=small  medium &miwxh=medium  large &miwxh=large  all wallpapers &miwxh=wallpaper  wide wallpaper &miwxh=widewallpaper  specific: &miwxh=640x480  &miwxh=800x600  &miwxh=1024x768  &miwxh=1152x864  &miwxh=1280x960  &miwxh=1280x1024  &miwxh=1600x1200  &miwxh=1792x1344  &miwxh=1800x1440  &miwxh=1856x1392  &miwxh=1920x1200  &miwxh=1920x1440  &miwxh=1600x600  &miwxh=2048x768  &miwxh=2560x1024  &miwxh=3200x1200 



  
Find: Photos    Graphics     Buttons/Banners

Color:    Sources:    Sizes:


Nemo's advice...

It seems that altavista image search and fast image search are searching the same database (yahoo bought altavista, fast and inktomi). Probably altavista have slightly broader 'SERPs' because it also searchs other sources: corbis, news and rollingstone.

Altavista image search has some interesting filters: you can filter out images that have the shape of Buttons or Banners (i.e. the ratio length/width is very different from 1). Altavista has also a filter to search graphics which is not too shabby, I think the algorithm consists in giving images whith a small color pallet.

Surprisingly (or maybe not:) altavista image search and fast image search share the same sintax, that is if you want to search for cat or dog you must type: (cat dog).

Altavista provides more images per pages and given what I said is a better choice than fast...

Altavista image search provides, for each image, a link with more info. There you'll find the text that was used to classify the image and the url of the image's page (some servers are configurated to show the images only if the correct reffering page is provided... that's a *big* problem with fast because it doesn't show the page's url). Basically the text used to classify the images are:
  • a snippet of text before or after the image
  • the anchor text on any link pointing to the image
  • the alt text of the image
  • the image's url
The match of keywords on the images name is a little more broader: a image named 'acat.jpg' will also match the keyword 'cat', the same is true for Google.

Google's algorithm is basically the same you still could see when Google was showing carelessly the image's cache... You can still see that Google is using the same strategy, but now you must work a little more to deduce it!

BTW this kind of algorithm has naturally implemented a proximity operator for the keywords *because* the snippets of text are usually quite small.
Nemo

ixquick Image search

June 2007: 60704 results: Not very good any more, alas...

Ixquick - Searches AltaVista, Fast Search, Excite, HotBot, Infoseek, MSN, Yahoo & more.
Web  News  MP3  Pictures 

Baidu Image search

The chinese google... here its "images" form

Baidu image search: main Baidu portal Quality form



FAST Image searching

FAST IMAGE SEARCH




FAST/alltheweb sorts its pictures in imagesearch (roughly) from larger to smaller and is therefore very useful when you are looking for large images, especially for a broad subject (especially when compared to google).

- ritz

Check also Nemo's advice for Altavista, above.

Excite image search
excite image search (draws from Fast  and Ditto)

 Excite Image Search 
Format  ALL  JPEG  GIF  BMP 
Type  ALL  COLOR  B/W  LINE ART 

For instance: snowflake

http://search-dyn.excite.co.uk/search.php?excitetype=image&external=1&maxCount=4&key=snowflake&ifa=on&ita=on
ifa = format all ~ ifj = format jpg ~ ita = type all ~ ifc = type color ~ etc.

Lycos image search
http://multimedia.lycos.com/ (Draws from Fast and adds and pushes commercial images from Getty

For instance: http://multimedia.lycos.com/results.asp?query=snowflakes&first=1&agree=1&component=MorePictures&sv=1:5|148E, Note the &agree=1 snippet, bypassing censorship.

Yahoo image search


For instance: http://images.search.yahoo.com/search/images/advanced?p=snowlakes

Show results withall of these words   
the exact phrase
any of these words
none of these words
Tip: Use these options to look for an exact phrase or to exclude pages containing certain words.
Size return images that are
Filetype return only image files formatted as
Coloration return only images in
Site/Domain
Tip: You can search for results in a specific website (e.g. yahoo.com) or top-level domains (e.g. .com, .org, .gov).
SafeSearch Filter On - Filter out mature images
 Off - Do not filter image results

   


Fotosearch image search


"Stock Photography and Footage The World's Stock Photography - One Web Site"
http://www.fotosearch.com/advancedsearch.asp: Very commercial oriented. Images are watermarked.


Freenet image search


German image search
http://www.freenet.de/suche/index.html?search_type=picture&adv=1

   Bildersuche 
 
  WO soll gesucht werden? (Where are we gonna search?)

Begriff muss als Text im Bild enthalten sein (Concept must be found as text inside the image)
Begriff muss im Dateinamen enthalten sein (Concept must be found inside the filename)
Im Bild muss das Gesicht einer Person zu sehen sein (Image must show a visage)
keine Werbebanner durchsuchen (Ditch advertisement banners)
  WAS soll gesucht werden? (What are we gonna search?)

Format nicht festgelegt (any format)        
JPEG (.jpg) GIF (.gif) BMP (.bmp)
             


Latin Guya image search


Now powered by Wanadoo.

This is a good search engine for spanish images.

Busca imagenes
 





Webshots image search


Another interesting search engine for images that is not bad at all, try the Advanced Search features.



Photo Search Advanced Search


Death of the "edu" trick

Image: snowflake, domain:edu
by fravia+

Contrary to the old wisdoms and promises of the ancients, the old trick of gong for quality limiting a search for images to the "edu" domains (as some obsolete search guides still underline), DOES NOT WORK ANYMORE, alas!
Please note the differences, both in quantity and in quality of the retrieved images:
This is google images search for snowflake: &q=snowflake (18900 good results)
Whereas...
this is google image search for snowflake limited to "edu" domains: &q=snowflake+site%3Aedu (688 bad results)


Samo samo with Fast/Alltheweb images search:
everything: q=snowflake (13399 good results)
versus
only edu: &q=snowflake+site%3Aedu (401 bad results)

The reason is that the very moment you specify a "Zusatz" to your query, the special algos that (try to) protect you from those snowflakes that are not snowflakish at all will disappear, just like a snowflake im hell. Hence you will be served with all "edu" results, indipendently from their non-pertinence to the original image query.
Alas, these results will be indipendent from their own crappiness as well: edu sites are not exclusive repositories of "a more scholarly context": on most "edu" domains (universities, colleges etcetera) students have PERSONAL PAGES (usually in some edu subdirectory beginning with a ~ tilde) where they gladly publish "le tout et n'importe quoi".

So if you want to "go for quality" you should rather change your arrows! Using for instance "snowcrystals" instead of "snowflakes" or trying some 'regional arrows' will give you rather different images.

Try for instance the following: &q=schneeflocke: note how google's help filters do not work for german, so you fish mucho noise among your signal, and note the difference between the previous search on US-google and the following "deutsche" search: schneeflocke.

Of course you can and should also try many more regional searches: &q=flocon.

Altosax's "not-image" trick

Turning a distraction in a search tecnique
by altosax

I was reading an interesting post on a newsgroup when i decided to search that argument on Google. I already had the Google page open so simply typed the words "shear modulus" and clicked the button. Well, what i had were only images. The fact was that the opened page was not www.google.com but images.google.com (that i also use). But this event gave me the idea to use the images search for everything else.

What i noticed infact was that a lot of sites were relevant for my search and for related arguments. This happens because when the words are searched in the pages content the hits returned also contain a lot of trash pages but how many pages display an image which name is related to your search?

This kind of search has a great advantage too, because you can visually choose which link to follow, because Google shows you the images that matched your search.

To better understand the point, try yourself my search
  1. on images_google: shear modulus
  2. and on good ole sites google: shear modulus...
Have you got the point?

altosax,
November 2002

Essays

  1. [tximasea.htm]: Of Lazyness, Greed, and Image Searching..., by -Tx: a very good 'image searching' essay, now (April 2000) accessible (it targeted a running context)
  2. [ima__sea.htm]: Searching an image without knowing its name (wizardry searches), by An Argy, part of the [essays]
  3. [ar1essay.htm]: The "index +of" seekers' trick (some web-searching knowledge, worked out and folded toghether)
    by aapje & ritz, part of the [essays].   "This query is used to find so called "open directories", by searching for certain standard-keywords that almost always appear in the server-generated pages of such directories..."
  4. [faf_task.htm]: Searching an image without knowing its name (wizardry searches), by Faf & Nemo, March 2003
  5. [ra_task.htm]: Solving an ochre colored "image searching" riddle (and challenge)
    by RA, June 2003

  6. [The importance of persistence: a case study]
    by Mordred and Jeff, part of the essays.htm & images.htm sections. "Images searching" Lore, March 2004

  7. A new short image riddle: [Entering a rough sea]
    by fravia+, July 2004
  8. [thefollowingrough.htm], by Thoughts, April 2006:
    it took almost TWO YEARS to solve the Entering a rough sea image riddle, part of the images searching lore section.


  9. [source2.htm]: Searching and understanding an image without knowing its name and purpose
    "The chick on the river", an autumnal, small, "image searching" assignement!
    by G. Bidwell and fravia+, november 2004
    Images finding on the web as art and passion.
    There are *many* hints in the image itself, will they be enough?
    Part of the finding images section and of the antiadvertisement section.

  10. More advertisement debunking, related to source2.htm.
    carocolo.htm: "What do we sell?"
    An easy one.
    There are (too) *many* hints in the image itself, will they be enough?

  11. More advertisement debunking, related to source2.htm.
    revenge.htm: "The head of the crab"
    Moderately hard.
    There are *many* hints in the image itself, will they be enough?

  12. More advertisement debunking, related to source2.htm.
    quietharbour.htm: "The quiet harbour"
    Hard one.
    There are *many* hints in the image itself, will they be enough?


  13. Content Based Image Recognition - a stab in PHP (Part 1)
    by Finn61, February 2006
    Part of the essays and of the images sections.
    Even the .edu's aren't giving all this research away. You will find some doors closed, although I was amazed at how easy it is to find info that people would like you to pay for, freely lying around, sometimes mistakenly, in other places. If you get a closed/pay database that allows you to preview pages of papers, then you know what to do to find the rest. ;)

  14. and

    Content Based Image Recognition - a stab in PHP (Part 2)
    by Finn61, February 2006
    Part of the essays and of the images sections.
    The design I finally decided to go with for image comparison was based on the frequency of colours in certain regions of the source and target images. I have chosen this design as a compromise for speed and experimentation. There are many variables to tweak, some values giving you good results and others terrible, but I wanted to build in flexibility so we can experiment and discover what works well for the sources and targets we have in mind.

  15. A new short image riddle: [Holy Holydays!]
    by fravia+, June 2006
  16. [searching_image_challenge_2008_1.htm]: A short image searching challenge for this cold February 2008
    by fravia+, January 2008
    part of the images section.
    "Well, how would you proceed?"


A discussion about finding images

How do you search for images? (26/04/01 15.17.40)
    Honestly , i have read everything fravia+ and company wrote on this but i still don't really understand which techniques you are supposed to use example , i recall a cherished image seen many many years ago somewhere i don't remember , i don't remember the name either , i don't know the artist , how do i proceed to find it????
GeeeG

Re: How do you search for images? (26/04/01 18.20.28)
    I've played around with image searching a lil bit and still have much to do to put any cohesion to any pattern that might work
    in your case you do not have any filename; nor, byte size, pixel size, or any alt statement to try to follow

    you must sit down and write up a list of everything you recall about the picture... you must repaint it with descriptive words (this does not mean you are going to find it; this is just how i would suggest you 'proceed to [try to] find it'

    what is the picture you remember? what is in the picture you remember?

    Build the picture with 'words'

    example: A picture of children jumping rope
    are they on school grounds? at home? in a meadow? whats going ON in the picture?

    some keywords might be: children kids jumping rope {children} playing school recess
    you must be careful which keywords you use in an engine such as googgle
    if you put children and kids in the same line and it does not FIND BOTH of those words on some exisiting web page then IT will NOT give you a return/ or it might find 10 hits with children and kids
    but find 1000 for chilren [without kids]
    or find 300 for kids [without children]
    When you have a GOOD idea of what you want to look for the less HITS is always better for sifting and narrowing down... but in your case the more hits the better because you do not know what you are looking for...why do i think it is better? well because by looking at the returns i skim and see if I can garner some better more meaningful list of words that connect to [children playground school jumping rope ect ect ect]
    what kind of dress style were these children in in the painting i remember?
    ohhhhhh very old time/ not modern/
    are they jumping rope on pavement? brick? dirt? gravel? these can be indicative of time periods also... ect

    How old could the painting be? go look up the history of rope jumping...[Im going to just make this up cause i don't have time to search it out] but lets say u search it and it comes back saying--- and the first instance of childreen jumping rope was because of this or that event and occurred in this or that country in 1734
    ok well now we know that unless your painter was orson wells who could conjure up children skipping rope that did not exist yet--that he was probably born after 1734...probably didn't start his starving artist carrer until he was 20 so probably an artist after 1754 plus+ or minus ... 300 years of artists doesn't narrow it down for you much...but you see where i am heading? at least you eliminated all paintings before 1734 :) Now what artists specialized in paintings of kids? [perhaps so; perhaps not--- don't get stuck on thinking he specialized (OR SHE!)] add art/artist/paintings of to word list...kindasorta...
    its hard to say because adding just ONE too many [as well as an incorrect one] words can completely change your return/s...
    whats in the background? forest rivers lakes mountains --- European? American?

    anyway i could go on and on ... the idea here is to BUILD a word list that describes your quarry

    this probably will explain it better:
    http://www.searchlores.org/ima__sea.htm

    good luck!!! good hunt!!!
rainbow

Re: Re: How do you search for images? (27/04/01 09.33.29)
    .oh well. some pointers here.

    I guess there are some questions you should ask yourself BEFORE beginning your searching task.

    1) Was that picture of a famous artist ? (whether or not you recall his name) like a painting or something alike ?
    2) Was it found in a site dealing with that sort of pictures (whether or not you can remember the url) or a commercial site?
    3)Was it of high quality ? (meaning over 200-300kbytes) or a small one? ..-100kb?

    I guess that if your answer to all above questions is something like:
    1) No
    2)No
    3)Small

    I'm sorry to say that you probably have NO chance of finding it (except if you are tooooooo lucky:).

    I dont have the time right now to explain why the above pointers (plus some others) are of great importance (but i will in a short kind of essay in the near future) but i've been dealing with the "how to find pictures" task a long time now.

    p.s Rainbows tactic is VERY CORRECT (grammatical anomaly here eheh) but, i think he will agree, that tactic will give you more possibilities into finding your picture if the answer on the above three questions is affirmative. His suggestions is THE way to navigate throught your results (after you have had a succesful query with small noise analogy that is) into extracting the picture you want (sort of like evaluating your results - results that you KNOW one of them should include your target).

    This post might seem a little fuzzy, but im on a deadline here eheh. I guess my pointers helped you a bit.

    I'll see you all later :)

axl


Loki's (quick) emperor's clothes

A 15.000 dollars question from our [messageboard]  :-)

Q:: Where does the image on searchlores itself, at the bottom of: Young slaves' behaviour, wabi, sabi and Levi's Jeans, (essay is called "How they exploit stupidity - part 1: The Emperor's New Clothes"), come from?
(http://www.searchlores.org/images/emperor2.jpg)


A:: Let's search.

google: emperor andersen

Page 2, that crown looks familiar

It's a cover of the book from the Starbright Fondation, as amazon told me. Published by Harcourt Brace & Co

In this bold and hilarious retelling, Hans Christian Andersen's classic fairy tale, The Emperor's New Clothes is re-imagined by an all-star celebrity cast. Among the writers are stars from the big and small screens, stage and music, as well as many other beloved personalities. Each celebrity contribution is illustrated by artists who have created some of the most treasured classics of American literature

http://www.starbright.org/emperors/bios/artists.html

That makes a lot of illustrators.. :)
I want the one who did the cover. I'm pretty sure it's the picture from fravia's essay. Same number of balls on the crown, same color, same shadows..

amazon allow to view some scanned pages from the book. let's have a look at the front flap

Jacket Illustration by William Joyce and Quentin Blake

let's fire google :
Quentin Blake : doesn't seem to be the artist that did the cover image. he probably did the front flap.

: looks better. more colours.

google :
william joyce emperor

bingo.


Just some comments about the previous search: some things valuable out of the pure 'technical' stuff.

once the image has a context, it's always easier to fish it out of the web, mostly because the image search engine index the pictures using the information they can gather on their pages (the 'physical' context).

usually, if you have only the picture, you need to build that context : making list of identified objects on the picture, colours, positions, subject, impressions etc.. There are plenty of methods to read and describe images. I'll try to find a good ressource.

Combined with the other concrete informations (size, format, filename..) you can think about a method, and forge your queries.

BUT, if the image is provided within a context, you can take some shortcuts. In the case of the emperor's clothes, once the connection with Andersen was made, most of the problem is solved. It's often the case with riddles, every detail count :)

I'm sure we could write/find/compile methods to read/describe/find images. But some may prefer the chaotic way. Thoughts ?

~

Another comment : we should take a closer look to amazon. it has more and more interesting features. the scanned pages of books are one, jeff and vvf found some audio oriented tricks, and recently they began to provide a 'full text search' for books..

loki


the original is here (11/11/03 17:19:59)
    Title (ID): The Emperor (WJO10A)
    Artist: William Joyce
    Source: The Emperor's New Clothes
    Image Size: 11 x 16 1/2 inches
    Medium: acrylic
    Price: $15000 (unframed)

    http://www.storyopolis.com/portfolio-dbp-detail.asp?ArtistID=135&ArtID=162




Images databases

Image Searching - What You See Is What You Get: Science Images on the Web (18/11/03 14:51:47)
    I was searching for ressources dealing with image description and analysis when i found some interesting ressources for the image searching section. A good addition would be to add a 'image databases' subsection, next to general search engines. Some of these databases may have a SE too.

    What You See Is What You Get: Science Images on the Web
    A selection of image-rich web sites in a variety of scientific disciplines is offered as a starting point for reference questions and educational programs. Tools for keeping up with new image resources are introduced. This review does not cover searching general World Wide Web sites or general commercial image databases for science images.

    Already reviewed, and classified ;)

    For evaluation purpose, here is the path i walked :

    Simple combing :) Gerverau is the author of a book i have, but in french, dealing with description and analysis of images. He is president of the International Association of Museums of History, curator of the Museum of Contemporary History (Paris) and director of the Cinema Museum (Paris). He wrote some scientific books on the subject of images and is the owner of imagesmag.net, a website dealing with research about image. So, my opinion was that he is a sort of authority in that domain :

    Searching on google : gervereau analysis images

    hit n8 : http://web.usal.es/~alar/Bibweb/Materias/I/imagenes.htm

    A commented bookmark, wich contains the WYSIWYG:SI ressources.

    That's it. I'll see if there are others gems in these, i'm still searching for good method of image analysis :)

    -----

    On a side note, still about images, i heard this morning in my 'IT and sociology' (i don't know how to name it) course about GETTY IMAGES and CORBIS.
    You (I) must have a look at what those company are in..

    More later.

loki

Images' semantics (evaluation lore)

When dealing with images (and eo ipso images' manipulation), in a world that not only allows, but actively encourages private ownership of the media, a sound knowledge of applied semantics, reality cracking and exegesis techniques may result quite useful...

Let's not forget (never!) that the 'bundling', the 'cut', the presentation and even the colors of an image possess tantamount importance for the message that the slavemasters want their readers to slurp. Reversers should always walk along +ORC's "thin cool line", and, for good measure always mistrust their own sources as well :-)
Brutality and compassion are BOTH presents in the above image, for instance. But it makes an helluja of a difference if you show the left or the right part of it :-(


On these matters see also the older essay Rhetoric of advertisement, a "Marlboro Classic" Advertisement analyzed



Please note that this section (searchlores' images.htm) is in progress, and that your own contributions, comments and hints are and will always be welcome: they represent the sine qua non in order to progress towards seeking perfection :-)
Petit image

© 3rd Millennium*: [fravia+], all rights reserved, reversed, revealed and reviled
*) Third millennium sounds a tag preposterous, uh, in fact circa 2000 out of 4,550,000,000 years... if you compress Earth's 4500 million years into the 24 hours timespan of a single day, humans will emerge from our monkeysh forefathers one minute and seventeen seconds before midnight, and our great "civilization" wont even enjoy that already short span.