MATLAB Answers

0

Image Processing: Do I Need to Train My Program with Sample Images for Machine Learning?

Asked by shiffraw dagnechaw on 11 Aug 2018 at 23:10
Latest activity Commented on by Florian Morsch on 13 Aug 2018 at 6:42

This is a broad question pertaining to image processing and coding. If I want to create a program in Matlab that enhances images in a automated fashion and( better contrast, edge sharpening etc. )that is specific to each image and changes them depending on a particular image basis and doesn't do the same exact enhancing for every image, would that be in the realm of machine learning and computer vision since the program changes the images accordingly? I'm a beginner coder and am working on this small project for a biophysics lab. They want me to write a program that takes two images of skin lesions ( skin tumors) and make a composite picture ( using subtraction method or division method). My only problem is that the function I've made enhances the images that I've been practicing with and probably won't enhance any arbitrary images given to me since different images have to be enhanced differently. So i'm not sure if I should learn advanced computer vision algorithims. Someone suggested to me that I would need about 1000 images to train my program. Is that true? Is there a way of creating a program that does this without needed images to train it?

  1 Comment

What is the specification on how they are enhanced? Like all pictures are taken from a different distance and in the end on every picture the lesions should have the same size? All pictures should have the contrast adjusted, so they all have the same contrast at the end? ) If so you dont necessarily need machine learning, you can do it in different ways.

You say better contrast and edge sharpening, i guess you picture quality changes and you want every outcome to be the same. So you have to find a base for every image and change the image according to that base. (Example, if you want to make the lesions all the same size, use a reference object when you take the picture. After that you could change the size according to the reference, so that every reference has the same size. Same thing can be done for other image processing elements)

Sign in to comment.

0 Answers