We present a family of adaptive Sobel filters for the geometrically correct estimation of the gradients of wide angle images. The proposed filters can be useful in a number of application domains exploiting wide angle cameras, as for instance, surveillance, automotive and robotics. The filters are based on Sobel's rationale and account for the geometric transformation undergone by wide angle images due to the presence of radial distortion. The proposed method is evaluated on a benchmark dataset of images belonging to different scene categories related to applications where wide angle lenses are commonly used and image gradients are often employed. We also propose an objective evaluation procedure to assess the estimation of the gradient of wide angle images. Experiments show that our approach outperforms the current state-of-the-art in both gradient estimation and keypoint matching.
We collected a dataset of 100 high resolution images belonging to the following scenes categories: indoor, outdoor, natural, handmade, urban, car, pedestrian, street. The considered categories are relevant to the main application domains where the image gradients are employed, and consistent with the scene categorization proposed by Torralba & Oliva. Each image is provided with one or more tags related to the above specified scene categories. All images have been acquired using a Canon 650D camera mounting a Canon EF-24mm lens and have resolution equal to 5204 X 3472 pixels.
This MATLAB code can be used to repeat the experiments and obtain the results presented in the paper.