Doubt- Image Processing

Hai Friends,
I’m having some basic doubts in image transforms.
• What is the difference between processing of original image and transformed image?
• If we are doing segmentation means , What is the difference between segmentation of original image and wavelet transformed image (subbands, for example approximation subband)?
• What is the need of transformation in processing of one image (for example classification , segmentation, etc)?
• Shall we improve the performance of the process by doing transformations?

hi
u seem to have lot of doubts in image processing.
transformation applied to a image helps in extracting the hidden features of the image. ex: consider a normal image,applying negative transformation to that image gives you a negative image in which u will be able to get the boundaries of that image.
processing of an original image may not help you in getting valuable information like low frequency components present in that image,but transforming an image help u in getting this informations.
segmentation of an original image helps u in segmenting that particular part u are in need of but segmenting an wavelet transformed image u will get a segmented image with good resolution and other regions of the image are also not affected.
u can just see the segmentation demo in matlb7 if u have it.and try to do wavelet based segmentation of the same image ur self u will see the difference in result.
mainly image processing algorthims what ever may be either transformation or segmentation is done to get the resolution of an image improved to the HVS - human visual system

Thank You very much friend,
I want to do image segmentation. By applying wavelet transform for one image, we apply segmentation for subbands. To get good results, we hav to reconstruct the original image. By doin this we can improve the performance. Am I right?
Once again thanks for ur help.

Does your method of img seg fall under any of these image seg classification?

Pixel based Segmentation
Model based Segmentation
Multi-scale Segmentation
Semi-automatic Segmentation

using a wavelet is a novel idea... how many stage of wavelet decomposition you are gonna use?