0

I need to develop the simple skin disease diagnose system using image processing and neural network. To use images in neural network, image should be same standard and, To identify the skin disease we should apply some image processing technique as well.

But I don't know what image processing technique apply first and what are they? As I read from references I think I need to apply image filtering technique, edge detection technique , etc...

Can someone who is expert in this please specify the image processing technique step by step. no need coding I just want to know image processing technique names and there order.

This is reference: reference

Gabriel Archanjo
  • 4,547
  • 2
  • 34
  • 41
user3565768
  • 43
  • 9

1 Answers1

0

I think when you say "images should be same standard" you mean that the images should represent similar patterns in similar form in order to train your neural network. However, I suppose you will not train your neural network using the image itself, in other words, the pixel values. You'll probably extract features from images for such purpose.

In the case you are dealing with patterns that can be recognized by its shape, in order to use "images in the same standard" you should apply morphological image operations in all images in your dataset and then extract shape features (area, width/height ratio, straight lines, angles, etc). These features could be the attributes input vector to your neural network and the pattern to be recognized could be the output.

The question is: what are the features necessary to properly identify the skin deseases? If you need to recognize spot in the skin, perhaps texture features. In order to extract texture features you will probably need to correct brightness variance in your dataset.

This is the just the tip of the iceberg. You could start choosing an image processing framework for such purpose:

The post below is a good start point:
"good resource for exploring image processing in java [closed]"

Community
  • 1
  • 1
Gabriel Archanjo
  • 4,547
  • 2
  • 34
  • 41