hhm@ihlpa.ATT.COM (Mayo) (08/12/87)
Where can I find information about the computerized enhancement of
black and white pictures to recover images not apparent on the original
print. How is it done? Is it applicable to film? What reference sources
are available?
Larry Mayo
AT&T Bell Laboratories
Naperville, Illthomson@udel.EDU (Richard Thomson) (08/13/87)
In article <5013@ihlpa.ATT.COM> hhm@ihlpa.ATT.COM (Mayo) writes: >Where can I find information about the computerized enhancement of >black and white pictures to recover images not apparent on the original >print. How is it done? Is it applicable to film? What reference sources >are available? > Larry Mayo I believe what you are referring to is called contrast enhancement. The basic idea is to take a histogram of the image and redistribute the small range in grey scale over the entire range from black to white. There is a very good reference on this technique as well as others for processing images using a computer. _Picture Processing_, (Azriel) Rosenfield & Kak This was the textbook in a course I took on image processing. Rosenfield is an old-timer in the field of digital image processing. He reviews articles for the IEEE Computer Graphics & Applications magazine and has published at least two editions of this book. The latest edition comes in two volumes, although the information on contrast enhancement is in the first volume. I don't have the second volume, so I don't know what's in it. :-( As far as its applicability to film, you would probably want to take an 'average' histogram to find a contrast expansion scheme that would uniformly increase the contrast in the whole film. I imagine that if you treated each frame as a separate photograph, light scenes would be darker and dark scenes would be lighter than desired Rich Thomson -- Rich Thomson Aspiring Grad Student ARPA: thomson@louie.udel.edu Forget Mars, we've got business on the Moon. UUCP: don't know OS/2: Yesterday's software tomorrow
jru@etn-rad.UUCP (John Unekis) (08/13/87)
In article <5013@ihlpa.ATT.COM> hhm@ihlpa.ATT.COM (Mayo) writes: >Where can I find information about the computerized enhancement of >black and white pictures to recover images not apparent on the original >print. How is it done? Is it applicable to film? What reference sources >are available? > The computerized enhancement of black and white pictures is a very broad and mature field and there exists thousands of volumes on the techniques involved. A brief summary would probably include the following steps - Image capture : The film or other source for the image is placed in front of a video camera or CCD array scanner and the image is reduced to an array of dots. The number of dots depends on the resolution of the device used, typical sizes are 640x480 (near TV quality picture), 512x512, or on higher resolution systems 1024x1024 or even 2048x2048. The higher the resolution, the more dots, and the better the image will look on the screen, but it will take correspondingly longer to process. The capture device feeds the intensity value for each dot to a device called an analog to digital converter, which assigns a numeric value to the intensity. The typical range for intensities is from 0 to 255 (Which happens to fit nicely in an 8-bit computer byte). The digitized image is then stored on computer disk or in computer memory as a two dimensional array of values. Image enhacement : Several types of image enhancement are possible. The most common, and most visually effective, is histogram equalization. If the original image is washed-out looking, where the black isn't very black, or the white isn't very white, or everything is just sort of gray, then if you look at the intensity values of the digital image they will use only a small portion of the available range. Out of the possible 0 to 255 , they may only use say, 80 to 120. One way to correct this is to calculate a histogram of the intensity values of the image. This would be an array with 256 locations, where each location would hold a count of how many dots in the image had that particular intensity value. Histogram equalization can be done by replacing each dot in the image with a value corresponding to the percentage of dot s in the image that are lower in value than this dot. Lets say that in our image a dot had the value 81, and that one percent of the dots in the image had values less than 81, we would the 81 with one percent of our available intensity, or about a 2. If another dot had a value of 120, and 99 percent of the dots in the image had values less than 120, then we would replace the dot with 99 percent of our possible intensity , or about 253. This process makes the black areas get very black, the white areas get very white, and in between areas have more visible contrast. An out of focus image can be made to look more in focus by a technique called edge enhancement. To do this we look at all the closest neighbors of a dot in a 3x3 area within the image , with the dot in question at the center. By replacing the dot in the center with 9 times itself, minus one times each of its eight neighbors, and by repeating this for each dot in the image, the picture will suddenly look sharper and more focused. One drawback is that this will exaggerate any graininess in the picture. There are dozens of other usefull techniques to enhace the digital version of images, far to many to go into here. Image display: A CRT display of the image that is being processed can be created using a display controller which converts digital intensity values for each dot of the image into analog intensities that can be shown on a display tube. Image Hardcopy: A new physical copy of the picture that was enhanced can be produced by sending the digital values for each dot in the image to a device such as a film recorder or laser printer which can produce dots of varying intensity on paper and thereby reproduce the intensity values of the image as a physical picture. More detailed documentation on all this can be obtained from text books on digital image processing which are available at any large university book store in the conputer science section. If you would like to actually buy equipment to do this sort of processing a simple low cost system for the IBM PC is available from Media Cybernetics by the title of 'IMAGE PRO'. More expensive, but higher speed and higher resolution systems are available from companies like VICOM of San Jose, CA. and Matrox of Ottawa Canada. If you have more specific questions, or would like to take a look at the image processing system that my company produces, email to me seismo!cbosgd!ihnp4!wlbr!etn-rad!jru or physical mail at John Unekis ms208 EATON corp. 31717 La Tienda Dr. Westlake Village, CA. 91359
hiebeler@csv.rpi.edu.UUCP (08/14/87)
Ah, that clarifies some things with me. I am doing some things with
cellular automata, such as diffusion-limited aggregation simulating
crystal formation, dust clustering, plant growth, etc. I had often
heard that automata was used for image analysis/processing/enhancement/
whatever. I had not really realized how this would be done, but what
was just said about 9xthe cell - each of its surrounding neighbors is
perfect for automata encoding. Maybe I'll run something like that on the
machine I've been using (a Cellular Automata Machine from MIT's lab,
it updates a 256x256 grid 60 times/sec, with each cell having 4 bits).
-D.H.
----
David Hiebeler hiebeler@csv.rpi.edu
Chatham, NY "Illusions, Richard! Every
(also Troy, NY) bit of it illusions!"hiebeler@csv.rpi.edu.UUCP (08/14/87)
References:
Keywords:
Ooops. That last was not too clear.
9xthe cell - its neighbors means
9 times the cell minus its neighbors.
That's better!
-D.H.
----
David Hiebeler hiebeler@csv.rpi.edu
Chatham, NY "Illusions, Richard! Every
(also Troy, NY) bit of it illusions!"lmiller@venera.isi.edu.UUCP (08/14/87)
In article <250@etn-rad.UUCP> jru@etn-rad.UUCP (0000-John Unekis) writes: >In article <5013@ihlpa.ATT.COM> hhm@ihlpa.ATT.COM (Mayo) writes: >>Where can I find information about the computerized enhancement of >>black and white pictures to recover images not apparent on the original >>print. How is it done? Is it applicable to film? What reference sources >>are available? >> > If you would > like to actually buy equipment to do this sort of processing a simple > low cost system for the IBM PC is available from Media Cybernetics > by the title of 'IMAGE PRO'. More expensive, but higher speed and > higher resolution systems are available from companies like VICOM > of San Jose, CA. and Matrox of Ottawa Canada. > > If you have more specific questions, or would like to take a look at > the image processing system that my company produces, email to me I can recommend highly the image processing software and imaging boards supplied by Werner Frei Work Phone: 213-452-1730 Company: Werner Frei Associates Work Address: 831 Pacific Street #1 Santa Monica, CA. 90405 Remarks: Publishes IMAGELAB software. We've used his image processing system on PCs at ISI for several years. Werner is an active researcher with a substantial publications record. His system is easy to install and use. Larry Miller USC/ISI
ted@mergvax.UUCP (William Klein) (08/26/87)
Another excellent reference book, used as a textbook in a course that
I took on digital image processing is:
DIGITAL IMAGE PROCESSING by Rafael C. Gonzalez/Paul Wintz
c. 1987 second edition, Addison-Wesley
It describes (in excruciating detail! :-( ) histogram enhancement, adaptive
transforms, noise removal, motion blur correction, homomorphic filtering,
etc, etc, etc. It is the best book I've seen on the topic.
Good luck, this is not trivial stuff to implement.
--
Life is what happens to you when you are busy making other plans.
Real Life: W. Ted Klein
UUCP: philabs!mergvax!ted
VOICE: 516-434-2687