Aug 10, 2015

Canny edge detector

The Canny edge detector, unlike convolutional edge detection filters, produces clear and thin lines when applied to an image.
The algorithm consists of four steps:
  1. Filter edges using a convolutional filter.
  2. Thin edges depending on it's gradient and direction.
  3. Threshold the image so all pixels where it's gradient is below a minimum threshold becomes black, all pixels where it's gradient is above a maximum threshold becomes white, and the rest becomes gray.
  4. Trace the edges so all gray pixels connected to white becomes white, and the rest becomes black.

Aug 6, 2015

Convolutional edge detection filters design

In this post I want to talk about edge detection filters, that are a type of convolution filters, this filters are widely used in features classification algorithms. I will show you all the theory beside edge detection and how to tweak the parameters to match your needs.

Jul 26, 2015

Denoise filters: Median

Gauss and Mean filters required a good knowledge of mathematics to understand its background. Median filter is very similar to both of them, but it based in a more logical approach than a mathematical one.

Jul 21, 2015

Denoise filters: Mean

Continuing with article series about the denoise filters, we had seen that the Gauss filter had a very high sensitivity to noise due to it's static nature. Now is time to analyze a dynamic denoise filter based in the arithmetic mean of the input pixels, this is called the Mean Denoise Filter. This filter adapts it's convolution kernel according to the values of the input pixels.

Jul 17, 2015

Denoise filters: Gauss

We have seen before about convolution filters, and at the end of that article we have seen a kernel used for removing noise in an image, that is random spurious pixels in a image, the process of removing that random pixels is called Denoise. In this case I will show you the general form of the denoising kernel based in gaussian distribution.