Adaptive enhancement of sea-surface targets in infrared images based on local frequency cues

Abstract

Image enhancement is an important preprocessing step of infrared (IR) based target recognition and surveillance systems. For a better visualization of targets, it is vital to develop image enhancement techniques that increase the contrast between the target and background and emphasize the regions in the target while suppressing noises and background clutter. This study proposes what we believe to be a novel IR image enhancement method for sea-surface targets based on local frequency cues. The image is transformed blockwise into the Fourier domain, and clustering is done according to the number of expected regions to be enhanced in the scene. Based on the variations in the elements in any cluster and the differences between the cluster centers in the frequency domain, two gain matrices are computed for midfrequency and high frequency images by which the image is enhanced accordingly. We provide results for real data and compare the performance of the proposed algorithm through subjective and quantitative tests with four different enhancement methods. The algorithm shows a better performance in the detail visibility of the target.

Adaptive enhancement of sea surface targets in infrared high dynamic range image

“…There is no standard or metric for the performance comparison of the enhancement techniques, so we experiment with subjective tests, observers are asked to give points to different images based on the four different criteria. The criteria are the truthfulness of the image, detail visibility of the target, unnatural artifacts, and total image quality [2] . Two examples for high dynamic range infrared images are given in Figure 3 and Figure 4.…”

“…To the best of our knowledge, Karali et al have developed an IR image enhancement technique specific for sea surface targets [2] [3] . But Karali et al's method is not able to adapt to different scenes and gives general performance with our image data.…”

Adaptive image enhancement based on clustering of wavelet coefficients for infrared sea surveillance systems

“…In our previous work [30], we proposed an adaptive enhancement method based on local frequency cues (AEMLFC) for sea-surface targets. The image is transformed blockwise into Fourier domain and clustering is done according to the number of expected regions to be enhanced in the scene.…”

“…In this study, we develop an adaptive enhancement method based on clustering of wavelet coefficients (AEMCWC). The work is different from [30] in a way that the proposed approach clusters wavelet coefficients to find regions to be enhanced and uses the relation between the subband images to find gains used in the reconstruction step. Another difference is that both clustering and enhancement operations are implemented in the wavelet domain, which makes the proposed method more robust when compared to that in our previous work where operations are implemented both in the spatial and frequency domain.…”

“…The compared algorithms are selected as AEMLFC, BCLAHE-CE, method of A are Mällo (MAM) [31], and a wavelet based enhancement method (WEM) [34]. The first three methods are chosen because they perform well in the subjective comparisons in our earlier work [30]. The last method is chosen for a fair comparison of the proposed method with another wavelet based method.…”