Previously we mentioned that computer vision is a way of enabling computers to see the world as we do but how can we help computers to improve their eyesight? Even though one way is to use more advanced algorithms there is another step that we can deploy which is called as preprocessing. Preprocessing allows us to improve visual quality of our data set which include images or video frames.

By default, AVT provides you with required tools to complete preprocessing step without needing to use any auxiliary products (Figure 1). Our preprocessing tool currently supports automatic contrast enhancement operations. In future, we will add functionalities like edge detection, corner detection etc.

With AVT you can either convert your images to grayscale or you can use them without changing their color properties. If you want to keep your images’ color properties, you can apply preprocessing in three different color spaces which are RGB, HSV, and LAB

For grayscale images, we provide three methods (Figure 2):

Histogram Equalization: Classical way of enhancing image visual quality that doesn’t require any user selected parameter to adjust image contrast.

Gamma Correction: If you wish to apply gamma correction, select the value of the "Gamma" adjustment parameter between 0.0 and 2.2.

CLAHE: This technique is a parameterized and adaptive version of the histogram equalization which allows you to fine tune your contrast adjustment process. It requires you to specify a "Clip Size" value, which is in the range 1-4, and also a "Tile Size" value, which represents an X-by-X square tile whose size is in the range 8x8, 16x16 and 32x32.

For color images, we provide four methods; same methods you can use for grayscale images, and a new addition to the list (Figure 3):

Sigmoid Correction: Similar to other parameterised methods, this method requires the parameters 'Gain (α)' and 'Cutoff (β)' to be specified. In applications, the "Gain" value is usually selected between 0.1 and 10, and the "Cutoff" value is selected between 0.0 and 1.0.

Summary

Preprocessing is an important step that can significantly improve the detection performance of your object detection application. At AVT we take this seriously, and we bring required tools to make your lives easier. All you require is just an internet connection and a web browser. We will take care of all other complex things for you.

If you choose AVT, you will be able to complete your preprocessing steps without needing anything extra!

Are you ready to start? Visit App