OpenCV image preprocessing for object detection

Question

I would like to know about some preprocessing practices recommended for applying to images before their features are computed, in order to increase the efficiency of the process as much as possible. (e.g. Croping, changing format, etc...)

Thanks in advance :)

fatihk · Accepted Answer · 2013-06-30 11:15:10Z

19

Preprocessing is heavily dependent on feature extraction method and input image type. Some common methods are:

Denoising: applying a Gaussian or simple box filter for denoising.
Contrast Enhancement: If gray level image is too dark or too bright, this may be applied.
Downsampling to increase speed.
Morphological operations for binary images.

edited Jun 30, 2013 at 11:15

answered Jun 30, 2013 at 10:03

fatihk

7,9491 gold badge28 silver badges51 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Str1101 Over a year ago

Thank you very much for your answer @thomas and for the great quality of the links you added, but now I have some doubts on how to apply these methods properly: 1- Should I detect the gray level of every image before applying the contrast enhancement or simply I apply it in all the cases? 2- What may be an appropriate size to the images? 3- Erosion, dilation, or both of them? (or none)

fatihk Over a year ago

@Str1101, 1- histogram equalization needs to be applied on gray level image 2- if you have some abnormal size like 1024x2048, any feature extraction process may take considerable time, so you can downsample such images by a factor of 2-3 and you can speed up this process upto 4-9 times, 3- Erosion can be used to filter out small regions in an image while dilation is used to remove some holes inside objects. They can also be used for RGB images. Erosion causes darker regions to widen while dilation has a reverse effect in RGB images.(as can be seen the cat image in the last link.)

Str1101 Over a year ago

Thanks again @thomas. The process has to be automated,so I think that I'll have to compare results to be able to determine which one of the morphological operations (and with which settings) works better.

Collectives™ on Stack Overflow

OpenCV image preprocessing for object detection

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related