Image Thumbnails That Represent Blur and Noise

Abstract

The information about the blur and noise of an original image is lost when a standard image thumbnail is generated by filtering and subsampling. Image browsing becomes difficult since the standard thumbnails do not distinguish between high-quality and low-quality originals. In this paper, an efficient algorithm with a blur-generating component and a noise-generating component preserves the local blur and the noise of the originals. The local blur is rapidly estimated using a scale-space expansion of the standard thumbnail and subsequently used to apply a space-varying blur to the thumbnail. The noise is estimated and rendered by using multirate signal transformations that allow most of the processing to occur at the lower spatial sampling rate of the thumbnail. The new thumbnails provide a quick, natural way for users to identify images of good quality. A subjective evaluation shows the new thumbnails are more representative of their originals for blurry images. The noise generating component improves the results for noisy images, but degrades the results for textured images. The blur generating component of the new thumbnails may always be used to advantage. The decision to use the noise generating component of the new thumbnails should be based on testing with the particular image mix expected for the application.

A no-reference metric for evaluating the quality of motion deblurring

“…Image quality metrics have been successfully applied in a variety of applications, such as automatic parameter selection [Zhu and Milanfar 2010;Mittal et al 2012a], blur-aware and noise-aware downsizing [Samadani et al 2010;Trentacoste et al 2011], and coded aperture design [Masia et al 2012]. We will discuss the difference between these methods and our approach in more detail in Sec.…”

“…Note that some previous approaches [Samadani et al 2010;Trentacoste et al 2011] also measure sharpness. However, Samadani et al [2010] use the shape of the PSF as its prior knowledge, and only Gaussian function is tested.…”

Quality-preserving image downsizing

“…Similar to our work, Samadani et al [2010] developed a method for amplifying specific artifacts present in full-size images to be visible in thumbnail images. However, their method cannot accurately detect motion blur or noise in textured areas.…”

Utilizing Image Scales Towards Totally Training Free Blind Image Quality Assessment

“…It has higher values for blur and noisy images. As shown in [5], these two distortions are greatly affected by the scale-space decomposition. Again, the standard deviation for the blurred and noisy images decreases, but for others the standard deviation increases.…”

“…The dissimilarity will increase with increase of distortion i.e more deviation from the natural image. It is already demonstrated in [5] that, standard image thumbnails that are generated by filtering and subsampling their originals lose similarity with them. Moreover, as discussed in [1], one of the motivations for developing multiscale oriented filters to decompose in frequency domain was partially inspired by the presence of neurons (in visual cortex) that are synchronized with visual signals of specific frequency, spatial frequency and orientation.…”