1458 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 20, NO. 5, MAY 2011
IMAGE Resolution Enhancement by Using Discrete andStationary Wavelet Decomposition
Hasan Demirel and Gholamreza Anbarjafari
In this correspondence, the authors propose an image resolu-tion enhancement technique based on interpolation of the high frequencysubband images obtained by discrete wavelet transform (DWT) and theinput image. The edges are enhanced by introducing an intermediate stageby using stationary wavelet transform (SWT). DWT is applied in orderto decompose an input image into different subbands. Then the high fre-quencysubbandsaswellastheinputimageareinterpolated.Theestimatedhigh frequency subbands are being modiﬁed by using high frequency sub-band obtained through SWT. Then all these subbands are combined togenerate a new high resolution image by using inverse DWT (IDWT). Thequantitative and visual results are showing the superiority of the proposedtechniqueovertheconventionalandstate-of-artimageresolutionenhance-ment techniques.
Discrete wavelet transform, image super resolution, sta-tionary wavelet transform.
Resolution has been frequently referred as an important aspect of animage. Images are being processed in order to obtain more enhancedresolution. One of the commonly used techniques for image resolutionenhancement is Interpolation. Interpolation has been widely used inmany image processing applications such as facial reconstruction ,multiple description coding , and super resolution –. There arethree well known interpolation techniques, namely nearest neighborinterpolation, bilinear interpolation, and bicubic interpolation.Image resolution enhancement in the wavelet domain is a relativelynew research topic and recently many new algorithms have been pro-posed –. Discrete wavelet transform (DWT)  is one of the re-cent wavelet transforms used in image processing. DWT decomposesan image into different subband images, namely low-low (LL), low-high(LH),high-low(HL),andhigh-high(HH).Anotherrecentwavelettransform which has been used in several image processing applica-tions is stationary wavelet transform (SWT) . In short, SWT is sim-ilar to DWT but it does not use down-sampling, hence the subbandswill have the same size as the input image.In this work, we are proposing an image resolution enhancementtechnique which generates sharper high resolution image. The pro-posed technique uses DWT to decompose a low resolution image intodifferentsubbands.Thenthethreehighfrequencysubbandimageshavebeen interpolated using bicubic interpolation. The high frequency sub-bands obtained by SWT of the input image are being incremented intothe interpolated high frequency subbands in order to correct the esti-mated coefﬁcients. In parallel, the input image is also interpolated sep-arately. Finally, corrected interpolated high frequency subbands andinterpolated input image are combined by using inverse DWT (IDWT)to achieve a high resolution output image. The proposed techniquehas been compared with conventional and state-of-art image resolu-
Manuscript received May 28, 2010; revised August 09, 2010; accepted Oc-tober 05, 2010. Date of publication October 18, 2010; date of current versionApril 15, 2011. The associate editor coordinating the review of this manuscriptand approving it for publication was Dr. Rafael Molina.H. Demirel is with the Department of Electrical and Electronic Engi-neering, Eastern Mediterranean University, Mersin 10, Turkey (e-mail:email@example.com).G. Anbarjafari is with the Department of Information Systems Engineering,CyprusInternationalUniversity,Mersin10,Turkey(e-mail:firstname.lastname@example.org).Digital Object Identiﬁer 10.1109/TIP.2010.2087767
tionenhancementtechniques.Theconventionaltechniquesusedarethefollowing:— interpolation techniques: bilinear interpolation and bicubicinterpolation;— wavelet zero padding (WZP).The state-of-art techniques used for comparison purposes are thefollowing:— regularity-preserving image interpolation ;— new edge-directed interpolation (NEDI) ;— hidden Markov model (HMM) ;— HMM-based image super resolution (HMM SR) ;— WZP and cycle-spinning (WZP-CS) ;— WZP, CS, and edge rectiﬁcation (WZP-CS-ER) ;— DWT based super resolution (DWT SR) ;— complexwavelettransformbasedsuperresolution(CWTSR).According to the quantitative and qualitative experimental results, theproposed technique over performs the aforementioned conventionaland state-of-art techniques for image resolution enhancement.II. P
In image resolution enhancement by using interpolation the mainloss is on its high frequency components (i.e., edges), which is due tothe smoothing caused by interpolation. In order to increase the qualityof the super resolved image, preserving the edges is essential. In thiswork, DWT has been employed in order to preserve the high frequencycomponents of the image. The redundancy and shift invariance of theDWT mean that DWT coefﬁcients are inherently interpolable .In this correspondence, one level DWT (with Daubechies 9/7 aswavelet function) is used to decompose an input image into differentsubband images. Three high frequency subbands (LH, HL, and HH)contain the high frequency components of the input image. In the pro-posed technique, bicubic interpolation with enlargement factor of 2 isapplied to high frequency subband images. Downsampling in each of the DWT subbands causes information loss in the respective subbands.That is why SWT is employed to minimize this loss.The interpolated high frequency subbands and the SWT high fre-quency subbands have the same size which means they can be addedwith each other. The new corrected high frequency subbands can beinterpolated further for higher enlargement. Also it is known that in thewavelet domain, the low resolution image is obtained by lowpass ﬁl-tering of the high resolution image . In other words, low frequencysubband is the low resolution of the original image. Therefore, insteadof using low frequency subband, which contains less information thanthe original high resolution image, we are using the input image forthe interpolation of low frequency subband image. Using input imageinstead of low frequency subband increases the quality of the superresolved image. Fig. 1 illustrates the block diagram of the proposedimage resolution enhancement technique.By interpolating input image by
, and high frequency subbandsby 2 and
in the intermediate and ﬁnal interpolation stages respec-tively, and then by applying IDWT, as illustrated in Fig. 1, the outputimage will contain sharper edges than the interpolated image obtainedby interpolation of the input image directly. This is due to the fact that,the interpolation of isolated high frequency components in high fre-quency subbands and using the corrections obtained by adding highfrequency subbands of SWT of the input image, will preserve morehigh frequency components after the interpolation than interpolatinginput image directly.III. R
Fig. 2 shows that super resolved image of Baboon’s picture usingproposed technique in (d) are much better than the low resolution
1057-7149/$26.00 © 2010 IEEE