Image-Processing Image Filters

Are there properties of digital images (e.g. dct coefficients, pixel values, YCbCr, others) that remain constant when filters like binarization, grayscale, sepia, etc, or tilting the image by a certain degree are applied. It would also be helpful if you could suggest any reading or online tutorial for basic image processing.

Upvotes: 0

Answers (3)

Sujan

Reputation: 1562

The properties of an image (i.e. pixels) always change when you process it. Processing simply means changing the pixel values in order to finally get something from it.

There are lots of image processing techniques such as removing noises, applying filters, re-sizing, cropping, edge detection, etc.

If you want to learn from the beginning then see tutorials of Bob Powell. It is in C# and quite easy to understand.

Upvotes: 0

Matt Montag

Reputation: 7490

It sounds like you want to know what features are robust to all sorts of image operations.

The properties you listed are not invariant to the transforms you listed. You ask if the "pixel values" remain constant when you apply a filter that by definition modifies the pixel values. The only positive answer about your list would be that DCT coefficients maintain their distribution when you apply a color filter.

I'm going to make an assumption and suggest that you should read up on feature detection, where the goal is to identify salient parts of an image that remain constant after a transformation like scaling, rotation, etc. These features are useful for image stitching, object detection, query-by-image search, and lots more.

Upvotes: 1