Evaluation methodologies provide a better understanding of the relationship between a technique and the image attributes. Metrics are used to evaluate the similarities between images. They may use different approaches depending on what needs to be achieved. If only objective values need to be compared statistics-based metrics are suitable. A number of image comparison metrics have been proposed in the literature that is based mostly on images statistics. This chapter shows the MATLAB code for the root mean square error (RMSE) calculation. It presents the MATLAB code for the computation of mean square error (MSE) between two images. Peak signal to noise ratio (PSNR) is another widely used metric, which takes into account the maximum value of the signal, and can be defined based on MSE. Metrics enable automation and the use of metrics can also be applied directly to methods, for example, to …

Large resolution High Dynamic Range (HDR) textures/images are becoming common in computer graphics applications. However, it is still a prohibitive task to perform on them a filtering step such as tone mapping in real-time. We introduce a novel technique called Selective Tone Mapper (STM) for accelerating the tone mapping step on large resolution HDR images. This technique presents an efficient memory usage, when compared with traditional methods, due to the reduced information required to localize the strong contrast areas.

The flexibility of current graphics hardware is still not enough to ensure the full implementation of an original complex algorithm such as a local tone mapping operator, which maintains the same quality performances as its original CPU implementation. Significant changes are often needed to the original CPU implementation in order to overcome many of the limitations of the current graphics hardware. As a result of this we often have reduced quality reproduction, and the frame rate of the GPU implementation is not always acceptable for real-time applications. In this paper, we show how to change the CPU implementation of a state of the art local tone mapping operator for accelerating the computation process to real time frame rates. We also present a modification of the luminance local adaptation computation, showing a simple but not yet exploited property of the Gaussian filter, allowing us to maintain the same quality appearance of the original tone mapping operator. Finally we test the hardware implementation on NVIDIA graphics cards on several images and as well as a video. We compare our hardware implementation with the corresponding CPU implementation and previous work.