Aplication of Multiple Filters over Images

Hey there! How are you all? Hope you’re healthy. At the moment of this post I am studying Digital Image Processing, under the tutelage of Agostinho Brito Jr. You can check out his online Digital Image Processing Classes by clicking on this link.

This post will account for a minor upgrade on the code filtroespacial.cpp. It is asked to apply two filters in sequence over a video stream: the laplacian and the gaussian.

I will highlight details that relates directly to my modifications. If you want to know more about how the original code works, please watch Agostinho’s explanation here.

A brief comment on how the original code works: after defining several filters, a video stream is opened (from a video camera). A loop lets the video be continuously processed, and the user can choose the filters using some keys: a - “absolute” filter;g - gaussian filter; h - horizontal sobel filter ; v - vertical sobel filter; l - laplacian filter; b - high boost filter. The processing loop works by making a filter available to be applied via switch-case and lately applying it over the frames.

How to create Laplacian of a Gaussian Processed Image?

Notice in the comment above that processing happens by making a filter available for application. The trick is to apply the first filter inside a case and making the next one available to applied in the loop.

To be clear how it works, consider the regular filter choosing (laplacian filter):

      case 'l':
        mask = cv::Mat(3, 3, CV_32F, laplacian);
        std::cout<<"filtro laplaciano:"<<std::endl;
        printmask(mask);
        cv::imwrite("./exercises_images/result_laplacian.png",result);
        break;

Now you can see the method described at the beginning of this topic:

      case 'p':
        mask = cv::Mat(3,3,CV_32F,gauss);
        std::cout<<"\t primeiro se aplica gauss:"<<std::endl;
        printmask(mask);
        cv::filter2D(frame32f, frameFiltered, frame32f.depth(), mask,
                     cv::Point(1, 1), 0);
        mask = cv::Mat(3, 3, CV_32F, laplacian);
        std::cout<<"\t e agora se aplica laplace:"<<std::endl;
        printmask(mask);
        cv::imwrite("./exercises_images/result_laplgauss.png",result);
        break;  

Another modification made was to save the image in the /exercises_images folder.

A Brief Comment in Laplacian-Gauss Filtering

During the tests, I couldn’t see any difference between applying the laplacian and the gaussian against applying just the laplacian. If you want to test itself, you can download the executable file here and save it on the /exercises folder. To run it just type:

./exercises/laplgauss  

Watch Out! I used python to convolute the gaussian and laplacian filters. I used the filter but the result was not good. If you want to play around with it, choose the convoluted filter using k.