Hello every body,

I want to make an application for a web cam:

The camera will capture 2 consecutive images of the scene and a java code i want to write must subtract the second image from the first so the result must be zero since they are the image of the same scene.

In order to subtract the 2 images from each other, I want to connect the camera to java and read the images (which are composed of pixels) as an array in java in order to subtract them cell by cell ( the array is composed of cells).

Any body can give me a hint in this step please ?

thank you

Maybe you find some useful information here:

http://www.javalobby.org/articles/ultimate-image/

Thanks, the link helped me a lot.

Do you know what is the command to subtract a pixel from another

This maybe is helpful: http://docs.oracle.com/cd/E17802_01/...escriptor.html

EDIT: By the way, the resulting image will almost never have all pixels exactly zero. Depending on the quality of your camera, more or less noise will be in the pictures that will lead to slightly different images.

You are right..while writing the code the result didn't give zero unless i loaded the same picture twice..My goal is that if the image changed it means that somebody entered the room so the alarm must turn on and a graph illustrating the trace of the intruder's movement is drawn and some other secure effects. I stick on the condition i want to write to make the camera realize the change of image by subtracting the two pictures from each other because even if the image didn't change the result of subtracting will not be zero. Any idea ?

May be motion is the thing you want.

The Motion depends on Linux which I am not familiar with. I get an idea, what if i changed the pictures taken by the camera to black and white before subtracting them. So if no one entered the room the result of subtracting will be zero ?

It doesn't matter if you convert the images before doing the subtraction. The noise is in the original images, so it also will be in any image that is based on them.

If noise is in each image taken of the scene how I will let the code assume they are similar.Is the code taking the same color of the scene different in the two images? If so how can I face this problem ? What if I make the condition to assume the two images the same is not that the RGB of the pixels are equal, but I make the condition that the RGB of the pixels of the first image minus the RGB of the pixels of the second image <5 for example ?

Yes, if you assume that noise is (if you have a somewhat decent cam) only a minimal change in the pictures, then ignoring minimal differences should work. It also may help to break down the images in smaller chunks. For example, if you have only a difference in one or two pixels in a 10x10 pixel block then it is unlikely that there was a motion, it is more likely that the differences are simply noise. The size of the pixel blocks should be dependent on the size of the original images. It wouldn't make much sense to use 50x50 pixel blocks with an image that is 320x240, but may be with an image that is 1920x1080.

It seems very improbable that your simplistic approach will ever be successful, due to a number of real-world factors. The imaging device will never be noise-free, and is subject to variations in sensitivity to various spectra due to temperature, possibly humidity, aging, noise coupled from power sources and noise from other electrical sources, etc. The optics of the imaging system will also be subject to long term changes. The image itself will change due to variations of lighting, mechanical vibrations of the imaging device, and other environmental factors.
To have any sort of reliable detection of movement, you will need to do more analysis than frame-by-frame comparison. You will probably need some form of image-recognition analysis, which is able to identify gross components within images, and then compare those on a frame-by-frame, or frame-series basis. There are packages available to assist with this, although I have no first-hand experience with any. I am given to believe that most higher quality packages are commercial products.
You may be able to glean some techniques from code that performs motion video compression. As I understand it, much of the compression technique is to compare frames, and then store only differences between sequential frames. At the root level, this sounds like what you are attempting to do, followed by some quantization of the frame-to-frame differences. There should be an abundance of open-source code that performs MPEG and other forms of motion video compression for your your scrutiny.

--- rod.