An advanced gradient histogram and its application for contrast and gradient enhancement

https://doi.org/10.1016/j.jvcir.2015.06.007Get rights and content

Highlights

  • We propose a histogram which contains intensity and gradient information of image.

  • The histogram can avoid the high peak essentially.

  • A novel image enhancement method for human vision is presented.

  • We prove in theory that the method can increase the gradient of image.

Abstract

This paper proposes an image contrast and gradient enhancement method based on advanced gradient histogram equalization (AGHE). We first define a novel advanced gradient histogram (AGH). Unlike the traditional intensity histogram which only contains intensity information, the AGH contains both gradient and intensity information of image. This character enables AGH to alleviate high peaks and thus avoid over enhancement in AGHE. Moreover, it’s proved that AGHE can increase the mean of absolute gradients (MAG) which is a measurement of image gradient. Then we present a sine function histogram correction (SHC) to control the enhancement level of AGHE. By modifying AGH using SHC before equalization, both the contrast and gradient enhancement levels can be controlled effectively. Simulation results demonstrate that AGHE with SHC (SAGHE) can improve the image subjective quality effectively by enhancing both the contrast and gradient of image.

Introduction

The contrast enhancement, as a kind of significant processing technique for both images and videos, can effectively improve the image visual quality for human perception and recognition. In addition, it is also an important preprocessing step to accentuate the essential features in images and videos for automatic pattern recognition, machine vision and other applications.

Various contrast-enhancement techniques have been developed. Histogram equalization (HE) is a widely used global enhancement method. Since the contrast gain is proportional to the height of the histogram, gray levels with larger pixel populations are expanded to a larger range of gray levels, whereas other gray-level ranges with fewer pixels are compressed to smaller ranges. Although HE can efficiently utilize display intensities, it tends to over enhance the image contrast if there are high peaks in the histogram, often resulting in a harsh and noisy appearance of the output image [1]. Numerous other global histogram equalization (GHE) methods have been proposed for limiting the level of enhancement, most of which are obtained through modifications on HE such as mean preserving bihistogram equalization (BBHE), brightness preserving histogram equalization with maximum entropy (BPHEME) and weighted thresholded HE (WTHE) [2], [3], [4], [5], [6], [7], [8]. These techniques usually outperform the basic HE technique. However, they fail to emphasize details of the local regions because they use histogram information over the whole image [5]. To overcome this limitation, local histogram equalization (LHE) based methods are developed [9], [10]. The LHE based methods generally require more computation and they not only highlight details in the image but also enhance noise.

Besides traditional histogram-based methods, there are also unconventional approaches to solve the contrast enhancement problem. Celik and Tjahjadi employ Gaussian mixture modeling (GMM) of an input image to perform nonlinear data mapping for enhancement [11]. The GMM is an automatic method and suitable for different types of images, however, it spends high computation cost. Recently, a contrast enhancement using adaptive gamma correction with weighting distribution (AGCWD) is proposed, which smoothes the fluctuant phenomenon by weighting distribution and enhances image automatically using gamma correction [12]. The AGCWD is straightforward while it may lose details in the bright regions of image when there are high peaks in the input histogram. Multiscale contrast enhancement techniques explore image decomposition before image enhancement to prevent artifacts. These techniques are computationally complex though generally yield high subjective quality [13], [14].

Whereas a variety of contrast-enhancement techniques have been proposed to improve the qualities of general images, all of them just utilize the intensity information of image without emphasizing the change of intensity. In fact, the human visual system (HVS) is more sensitive to intensity changes, i.e., the gradient than the absolute intensity of image [15], [16], [17], [18]. Moreover, the enhancement of gradient is essential in many subsequent applications of image enhancement such as target detection and recognition considering the high gradient pixels to be the edge which is a basic feature of target [19], [20]. Our goal in this paper is to develop an image enhancement algorithm, which is capable of enhancing image contrast in general and enhancing the gradient specially. The main contributions of this paper are:

  • Propose a novel advanced gradient histogram (AGH). Besides pixel population information of each intensity, AGH contains gradient information of image, which reflects the range of intensity levels at which details of an image occur. Moreover, the peaks of AGH are much lower than those of traditional histogram which only contains intensity information. This can alleviate the over-enhancement phenomenon of enhancement algorithms based on traditional histogram.

  • Propose a straightforward and visually pleasing image enhancement method, i.e., SAGHE. To the best of our knowledge, SAGHE is the first image enhancement technique which can enhance both the contrast and gradient of image. The sine function histogram correction (SHC) used to modify AGH before equalization can control the enhancement levels of both contrast and gradient strength effectively using a parameter n.

  • Define mean of absolute gradients (MAG) as a measurement of image gradient. Give a universal expression of MAG of enhanced images using GHE techniques and further prove that the theoretical maximum value of MAG of GHE techniques can be achieved using our SAGHE with n=- in SHC.

The rest of this paper is organized as follows: Section 2 presents a detailed description of the proposed AGH. Section 3 proposes the novel image enhancement method SAGHE. Section 4 presents experimental results. Finally, Section 5 concludes this paper.

Section snippets

The advanced gradient histogram

In this section, we first give the definition of the mean of absolute gradients (MAG) and then give a detailed description of the proposed advanced gradient histogram (AGH).

The AGHE with SHC

In this section, we will introduce the proposed AGHE with SHC (SAGHE). The main steps of our SAGHE are given in Fig. 3. We first obtain the AGH of input image according to the method introduced in the last section. Then, present a sine function histogram correction (SHC) theme to modify the AGH. Finally, apply histogram equalization to the AGH after SHC and obtain the enhanced image. If there is no SHC, our method just equalizes the new AGH. We prove theoretically in this section that the MAG

Results and discussion

In this section, we provide experimental results in order to demonstrate the effectiveness of the proposed SAGHE in comparison to the conventional HE, a recently proposed contrast enhancement method using adaptive gamma correction with weighting distribution (AGCWD) [12] and a contrast enhancement based on layered difference representation of 2D histograms (LDR) [22]. Shih-Chia Huang et al. compare AGCWD against some state-of-the-art algorithms proposed in [6], [7], [24], [25], [26] and show

Conclusion

In this paper, we presented a novel enhancement method for human vision, which can enhance both the gradient and contrast of image and thus produces a high subjective quality result. The proposed algorithm is based on an originally proposed advanced gradient histogram (AGH). Unlike the traditional histogram which only has intensity information, the AGH contains both intensity and gradient information of image, which can alleviate the contrast overstretching, especially when there is high peak

References (32)

  • C. Lee et al.

    Power-constrained contrast enhancement for emissive displays based on histogram equalization

    IEEE Trans. Image Process.

    (2012)
  • J.-Y. Kim et al.

    An advanced contrast enhancement using partially overlapped sub-block histogram equalization

    IEEE Trans. Circ. Syst. Video Technol.

    (2001)
  • T. Celik et al.

    Automatic image equalization and contrast enhancement using Gaussian mixture modeling

    IEEE Trans. Image Process.

    (2012)
  • S.-C. Huang et al.

    Efficient contrast enhancement using adaptive gamma correction with weighting distribution

    IEEE Trans. Image Process.

    (2013)
  • W.-F. Lee et al.

    Perception-based high dynamic range compression in gradient domain

  • R. Fattal

    Edge-avoiding wavelets and their applications

    ACM Trans. Graph. (TOG)

    (2009)
  • Cited by (16)

    • Crack identification method of highway tunnel based on image processing

      2023, Journal of Traffic and Transportation Engineering (English Edition)
    • TEDLESS – Text detection using least-square SVM from natural scene

      2020, Journal of King Saud University - Computer and Information Sciences
      Citation Excerpt :

      Other eminent researchers who have worked in pre-processing the image are Huang and Lin. Huang et al. propose a system to enhance the contrast based on advanced gradient histogram equalization and controls the enhancement of contrast using Sine histogram correction (Huang et al., 2015). Lin et al. have used a pipelined approach has been followed to correct color distortion by histogram averaging and histogram remapping (Lin et al., 2015).

    • Image contrast enhancement based on intensity expansion-compression

      2017, Journal of Visual Communication and Image Representation
      Citation Excerpt :

      A simpler implementation operation was proposed in [9], where a dynamic stretching scheme is used instead of employing the weighted sum strategy. In [10], an alternative target histogram construction method, that makes use of local edge features together with the pixel intensities, was developed. An alternative definition of the target histogram was suggested in [11], where an iterative process was employed to match a uniform distribution and preserve the original histogram shape.

    • Image contrast enhancement using fuzzy clustering with adaptive cluster parameter and sub-histogram equalization

      2017, Digital Signal Processing: A Review Journal
      Citation Excerpt :

      The last step contains a weighted factor to obtain an average from the original and the enhanced image, to control possible over enhancement. Huang developed AGHE to deal with the problem of available brightness saturation, arising when working with HE method [30]. Brightness saturation problem is seen in areas of an image in which the corresponding histogram is narrow, which causes constant alterations in the transform function, resulting in brightness saturation.

    View all citing articles on Scopus

    This paper has been recommended for acceptance by Yehoshua Zeevi.

    View full text