1,721,100 research outputs found
Metodi non-lineari per la cancellazione del rumore in dati di misura 1-D e 2-D: validazione dei risultati
No full text
New Method for Measuring the Detail Preservation of Noise Removal Techniques in Digital Images
Full text linkIt is known that cancelling the noise without blurring the image details is a very difficult task for any image denoising technique. The availability of metrics for accurate evaluation of filtering distortion is thus of paramount importance for the development of new filters. Peak signal-to-blur ratio PSBR is a recently introduced measure of detail preservation that overcomes the limitations of the sole peak signal-to-noise ratio (PSNR) and other metrics in evaluating the performance of image denoising filters. Formally, the PSBR is the PSNR component that deals with the detail blur, so the method that is adopted for blur estimation plays a key role. This paper presents a novel algorithm for PSBR computation that offers significant advantages over the first method: it is simpler, more robust and much more accurate. Furthermore, this paper presents new validation tools for evaluating the accuracy of this kind of metrics when some well known classes of linear
and nonlinear filters are considered. Results of many computer simulations dealing with images corrupted by different combinations of Gaussian and impulse noise show that the proposed PSBR algorithm outperforms the most effective metrics in the field
Automatic Enhancement of Noisy Images Using Objective Evaluation of Image Quality
No full textA new method for automatic enhancement of noisy images is presented. The most relevant feature of the proposed approach is a novel procedure for automatic parameter tuning that takes into account the histograms of the edge gradients. Thus a satisfactory compromise between smoothing and sharpening can be automatically found according to the human perception. The proposed method is favorably compared to other techniques in the literature
A Method Based on Piecewise Linear Models for Accurate Restoration of Images Corrupted by Gaussian Noise
No full textPiecewise linear (PWL) models are very attractive for image processing due to their simplicity and effectiveness. A new filtering architecture adopting multiparameter PWL functions is proposed for accurate restoration of images corrupted by Gaussian noise. The filtering performance is analyzed by taking into account the different behavior from the point of view of noise removal and detail preservation. The sensitivity to a change of the parameter settings is also investigated. In the new approach, the parameter values are automatically selected by resorting to a procedure that estimates the standard deviation of the Gaussian noise. Results dealing with different test images and noise variances show that the method yields a very accurate restoration of the image data
Study of the accuracy of the color peak signal-to-blur ratio (CPSBR)
No full textThe most critical issue in the restoration of color pictures from noise is the preservation of the useful information embedded in the image data. The Color Peak Signal-to-Blur Ratio (CPSBR) is a new full-reference method that measures the color/detail preservation yielded by a color image denoising filter. The approach is based on a simple and effective algorithm for the estimation of the filtering blur that operates in the RGB color space. An extensive study of the accuracy of the CPSBR is provided in this paper focusing on two key paradigms for image denoising: the family of order-statistics smoothers and the class of nonlinear weighted average filters. In this framework, the exact values of color distortion and detail blur produced by weighted vector medians, scalar and vector bilateral filters are theoretically evaluated and used for a comparison in order to validate the method. Results of many computer simulations dealing with color images corrupted by different amounts of Gaussian and impulse noise show that the novel CPSBR is a very accurate measure of color/detail preservation
- …
