CN107169925A - The method for reconstructing of stepless zooming super-resolution image - Google Patents

Abstract

The invention relates to a reconstruction method of a steplessly zoomed super-resolution image, comprising three processing steps: a dictionary training process, an image reconstruction process, and a scale change process. The invention proposes a method for establishing an image database training dictionary, trains the dictionary database off-line with the image set, obtains the mapping relationship between high-resolution images and low-resolution images, optimizes the image reconstruction steps, effectively reduces the running time of the method, and is self-adaptive and efficient sex. The threshold value in the method of the present invention is optimized and fixed based on a large number of comparative experiments, and does not need to be changed during the process, effectively improving the image reconstruction efficiency without affecting the image reconstruction effect. The super-resolution reconstructed image obtained by the present invention can be subjected to scale transformation of any size. In engineering applications, it can meet the needs of different situations and has good applicability.

Description

A Reconstruction Method for Stepless Scaling Super-resolution Images

technical field

The invention belongs to the technical field of image reconstruction, and in particular relates to a reconstruction method of a steplessly zoomed super-resolution image.

Background technique

At present, the image resolution will be affected by factors such as the imaging system and the shooting environment, and the obtained image resolution often cannot meet people's needs. However, the current optical imaging system and its detector are limited by the technical level and processing complexity, and it is impossible to improve the imaging quality of the system from the perspective of hardware. Therefore, the emergence of image super-resolution reconstruction technology breaks the limitation of image resolution. Through software processing, a high-resolution image is calculated from one or more degraded low-resolution images to obtain richer detail information. At present, super-resolution reconstruction technology is widely used in the fields of image processing and optical imaging, and has attracted people's attention.

At present, the learning-based sparse representation image super-resolution reconstruction algorithm can effectively improve the resolution of the image, and the scale of the image is enlarged with a magnification of 3. However, in some specific cases, the high-resolution images obtained by this method cannot meet the required scale requirements.

Therefore, how to provide an image reconstruction method so that the obtained super-resolution reconstructed image can be transformed at any scale to meet the relevant industrial requirements becomes very important.

Contents of the invention

In order to solve the above-mentioned problems existing in the prior art, the present invention provides a kind of reconstruction method of stepless scaling super-resolution image, comprising the following steps:

Step 1, blurring and downsampling the high-resolution sample image according to the degradation model to obtain a low-resolution sample image;

Step 2, using the K-SVD method to perform dictionary training using the low-resolution sample image to obtain a high-resolution dictionary and a low-resolution dictionary;

Step 3, preprocessing the low-resolution image to be reconstructed to obtain several low-resolution image blocks;

Step 4, using the low-resolution sample image to calculate a fixed threshold;

Step 5, judging whether the pixel value of each of the low-resolution image blocks is smaller than the fixed threshold; if so, then judging that the low-resolution image block is a low-information image block; Low-resolution image blocks are high-information image blocks;

Step 6: Reconstruct the high-information image block using a dictionary learning and sparse representation algorithm to obtain a high-information reconstruction sub-region;

Step 7, reconstructing the low-information image block using an interpolation algorithm to obtain a low-information reconstruction sub-region;

Step 8, performing image stitching on the high-information quantum region and the low-information reconstruction sub-region to obtain a super-resolution reconstruction image;

Step 9: Perform M-fold interpolation on the super-resolution reconstructed image using an interpolation algorithm to obtain an N*M (N=2, 3, 4; M>0)-fold super-resolution reconstructed image.

In one embodiment of the present invention, said step 2 includes the following steps:

(21) Extracting image features of the low-resolution sample image by a feature extraction method to obtain high-resolution feature information and low-resolution feature information of the space object.

(22) Using the K-SVD method, perform joint training on the high-resolution feature information and the low-resolution feature information to obtain the high-resolution dictionary and the low-resolution dictionary.

In one embodiment of the present invention, described step (22) comprises the following steps:

(221) Utilize sparse K-SVD method to train low-resolution dictionary;

(222) Compute a high-resolution dictionary.

In one embodiment of the present invention, said step 3 includes the following steps:

(31) Denoising the low-resolution image to be reconstructed to obtain a first image;

(32) Deblurring the first image to obtain a second image;

(33) Segment and save the second image according to a fixed length and width to form the low-resolution image block.

In one embodiment of the present invention, said step 4 includes the following steps:

(41) performing block segmentation on the low-resolution sample image to obtain a low-resolution sample image block;

(42) Using an edge extraction algorithm to extract edge information of the low-resolution sample image blocks, counting the pixel values of each of the low-resolution sample image blocks and the distribution of pixel values of all the low-resolution sample image blocks;

(43) Selecting X (X>1) candidate thresholds according to the pixel values of each of the low-resolution sample image blocks and the distribution of pixel values of all the low-resolution sample image blocks;

(44) Utilizing a learning-based sparse representation image super-resolution reconstruction algorithm for each of the candidate thresholds to form a plurality of high-resolution candidate images;

(45) Select the information amount threshold from X candidate thresholds according to the calculation time of the learning-based sparse representation image super-resolution reconstruction algorithm and the resolution of the correspondingly formed high-resolution candidate image.

In one embodiment of the present invention, the edge extraction algorithm is a Canny operator edge detection algorithm.

In one embodiment of the present invention, described step 6 comprises the following steps:

(61) performing image filtering processing on the high-information image block, and performing high-frequency feature extraction to obtain high-frequency information;

(62) Using the high-frequency information to perform KPCA dimensionality reduction on the high-information image block to realize high-dimensional data compression;

(63) Using the OMP algorithm to reconstruct the data after dimensionality reduction and compression to obtain the reconstruction sub-region with high information content.

In one embodiment of the present invention, the image filtering operation uses a filter bank of two-dimensional filtering operators, and the filter bank is f={f ₁ , f ₂ , f ₃ , f ₄ }, wherein,

f ₁ =[1,-1], f ₂ =f ₁ ^T

f ₃ =LOG, f ₃ =f ₃ ^T

T is a matrix transpose operation, and LOG is a 5×5 two-dimensional filter operator.

In one embodiment of the present invention, described step (63) comprises the following steps:

(631) Calculate the sparse representation coefficient β according to the OMP algorithm, wherein,

Among them, y is the data after dimension reduction and decompression, T ₀ is a given sparsity, and β _i is a sub-element in matrix β;

(632) Multiply the sparse representation coefficient β by the high-resolution dictionary to obtain the high-information reconstruction sub-region.

In one embodiment of the present invention, the interpolation algorithm is bicubic interpolation algorithm.

The embodiments of the present invention have the following beneficial effects,

1. The embodiment of the present invention proposes a method for establishing a training dictionary in an image library, trains the dictionary library offline with an image set, obtains the mapping relationship between a high-resolution image and a low-resolution image, optimizes the image reconstruction steps, effectively reduces the running time of the method, and has automatic adaptability and efficiency.

2. The super-resolution reconstructed image obtained by the present invention can be scaled to any size. In engineering applications, it can meet the needs of different situations and has good applicability.

3. The threshold value in the method of the present invention is optimized and fixed based on a large number of comparative experiments, and does not need to be changed during the process, and the image reconstruction efficiency is effectively improved without affecting the image reconstruction effect.

4. Scaling is performed based on the super-resolution reconstructed image. Since the algorithm reconstructs high-frequency information, the zoom effect will be better.

Description of drawings

FIG. 1 is a schematic flowchart of a method for reconstructing a steplessly zoomed super-resolution image provided by an embodiment of the present invention;

Fig. 2 is a schematic diagram of a dictionary training principle provided by an embodiment of the present invention;

FIG. 3 is a schematic diagram of a method for reconstructing a steplessly zoomed super-resolution image provided by an embodiment of the present invention.

detailed description

The present invention will be described in further detail below in conjunction with specific examples, but the embodiments of the present invention are not limited thereto.

Embodiment one

Please refer to Fig. 1, Fig. 2 and Fig. 3, Fig. 1 is a schematic flow chart of a method for reconstructing a stepless zoom super-resolution image provided by an embodiment of the present invention; Fig. 2 is a schematic diagram of a dictionary training principle provided by an embodiment of the present invention ; FIG. 3 is a schematic diagram of the principle of a method for reconstructing a steplessly zoomed super-resolution image provided by an embodiment of the present invention. The reconstruction method of the steplessly zoomed super-resolution image comprises the following steps:

Step 1, blurring and downsampling the high-resolution sample image according to the degradation model to obtain a low-resolution sample image;

Step 2, using the K-SVD method to perform dictionary training using the low-resolution sample image to obtain a high-resolution dictionary and a low-resolution dictionary;

Among them, step 1 and step 2 establish the method of image library training dictionary, the image set is trained offline to obtain the mapping relationship between high-resolution images and low-resolution images, optimize the image reconstruction steps, effectively reduce the running time of the method, and have automatic adaptability and efficiency.

Step 3, preprocessing the low-resolution image to be reconstructed to obtain several low-resolution image blocks;

Step 4, using the low-resolution sample image to calculate a fixed threshold;

Step 5, judging whether the pixel value of each of the low-resolution image blocks is smaller than the fixed threshold; if so, then judging that the low-resolution image block is a low-information image block; Low-resolution image blocks are high-information image blocks;

Among them, step 5 significantly improves the reconstruction speed while ensuring the reconstruction quality, which is suitable for applications with high time performance requirements.

Step 6: Reconstruct the high-information image block using a dictionary learning and sparse representation algorithm to obtain a high-information reconstruction sub-region;

Among them, the learning-based sparse representation reconstruction method avoids the artificial selection of the number of neighbors, and the reconstruction effect is better.

Step 7, reconstructing the low-information image block using an interpolation algorithm to obtain a low-information reconstruction sub-region;

Step 8, performing image stitching on the high-information quantum region and the low-information reconstruction sub-region to obtain a super-resolution reconstruction image;

Step 9: Perform M-fold interpolation on the super-resolution reconstructed image using an interpolation algorithm to obtain an N*M (N=2, 3, 4; M>0)-fold super-resolution reconstructed image.

Specifically, the step 2 includes the following steps:

(21) Extracting image features of the low-resolution sample image by a feature extraction method to obtain high-resolution feature information and low-resolution feature information of the space object.

(22) Using the K-SVD method, perform joint training on the high-resolution feature information and the low-resolution feature information to obtain the high-resolution dictionary and the low-resolution dictionary.

Wherein, described step (22) comprises the steps:

(221) Utilize sparse K-SVD method to train low-resolution dictionary;

(222) Compute a high-resolution dictionary.

Specifically, the step 3 includes the following steps:

(31) Denoising the low-resolution image to be reconstructed to obtain a first image;

(32) Deblurring the first image to obtain a second image;

(33) Segment and save the second image according to a fixed length and width to form the low-resolution image block.

Specifically, said step 4 includes the following steps:

(41) performing block segmentation on the low-resolution sample image to obtain a low-resolution sample image block;

(42) Using an edge extraction algorithm to extract edge information of the low-resolution sample image blocks, counting the pixel values of each of the low-resolution sample image blocks and the distribution of pixel values of all the low-resolution sample image blocks;

(43) Selecting X (X>1) candidate thresholds according to the pixel values of each of the low-resolution sample image blocks and the distribution of pixel values of all the low-resolution sample image blocks;

(44) Utilizing a learning-based sparse representation image super-resolution reconstruction algorithm for each of the candidate thresholds to form a plurality of high-resolution candidate images;

(45) Select the information amount threshold from X candidate thresholds according to the calculation time of the learning-based sparse representation image super-resolution reconstruction algorithm and the resolution of the correspondingly formed high-resolution candidate image.

Preferably, the edge extraction algorithm is a Canny operator edge detection algorithm;

Furthermore, said step 6 includes the steps of:

(61) performing image filtering processing on the high-information image block, and performing high-frequency feature extraction to obtain high-frequency information;

(62) Using the high-frequency information to perform KPCA dimensionality reduction on the high-information image block to realize high-dimensional data compression;

(63) Using the OMP algorithm to reconstruct the data after dimensionality reduction and compression to obtain the reconstruction sub-region with high information content.

Further, the image filtering operation adopts a two-dimensional filtering operator filter bank, and the filter bank is f={f ₁ , f ₂ , f ₃ , f ₄ }, wherein,

f ₁ =[1,-1], f ₂ =f ₁ ^T

f ₃ =LOG, f ₃ =f ₃ ^T

T is a matrix transpose operation, and LOG is a 5×5 two-dimensional filter operator.

Further, said step (63) includes the following steps:

(631) Calculate the sparse representation coefficient β according to the OMP algorithm, wherein,

Among them, y is the data after dimension reduction and decompression, T ₀ is the given sparsity, and β _i is the sub-element in the matrix β.

(632) Multiply the sparse representation coefficient β by the high-resolution dictionary to obtain the high-information reconstruction sub-region.

Preferably, the interpolation algorithm is bicubic interpolation algorithm. It is also possible to employ, for example, linear interpolation, nearest neighbor method, bilinear interpolation method, and the like.

This embodiment has the following advantages:

1. The embodiment of the present invention proposes a method for establishing a training dictionary in an image library, trains the dictionary library offline with an image set, obtains the mapping relationship between a high-resolution image and a low-resolution image, optimizes the image reconstruction steps, effectively reduces the running time of the method, and has automatic adaptability and efficiency.

2. The super-resolution reconstructed image obtained by the present invention can be scaled to any size. In engineering applications, it can meet the needs of different situations and has good applicability.

3. The threshold value in the method of the present invention is optimized and fixed based on a large number of comparative experiments, and does not need to be changed during the process, and the image reconstruction efficiency is effectively improved without affecting the image reconstruction effect.

Embodiment two

On the basis of the first embodiment above, this embodiment provides another stepless zoom super-resolution image reconstruction method, which includes three processing steps: a dictionary training process, an image reconstruction process, and a scale change process. Specifically include the following steps:

S1: Dictionary reconstruction process.

S11: Using a large number of high-resolution sample images, the high-resolution images are blurred and down-sampled by N times according to the corrected degradation model to obtain corresponding low-resolution sample images.

S12: Extract image features from the low-resolution sample image obtained in step 1 through a feature extraction method, and obtain high and low-resolution feature information of the space object, namely X _s and Y _s .

S13: Using the K-SVD method, perform joint training on the high and low resolution feature information to obtain a high and low resolution dictionary.

S13a: Training a low-resolution dictionary. The base dictionary Φ selects an over-complete DCT dictionary, and uses the sparse K-SVD method to solve:

Then the low-resolution dictionary D _l =ΦW, W is an atom representation matrix. Compared with the parsing dictionary model, the double-sparse dictionary model provides adaptability through the modification of W.

S13b: Calculate a high-resolution dictionary. Assuming that the high-resolution-low-resolution image blocks have the same sparse representation coefficient A under the high-resolution-low-resolution dictionary pair, the high-resolution dictionary D _h can be calculated by minimizing the approximation error in the following formula:

Solve using the pseudoinverse:

D _h ＝X _s A ⁺ ＝X _s A ^T (AA ^T ) ^-1 (3)

Among them, the superscript "+" indicates the pseudo-inverse.

S2 image reconstruction process.

S21: Perform preprocessing on the low-resolution image to be reconstructed, which mainly includes image denoising, image deblurring, and sample block operations. Its processing steps are:

S21a: Denoising the low-resolution image to be reconstructed;

S21b: deblurring the image obtained in step S21a;

S21c: dividing the image obtained in step S21b into blocks, dividing and saving the entire image according to a fixed length and width;

S22: Using the low-resolution sample image obtained in step S11 to calculate a fixed threshold, the specific processing steps are as follows:

S22a: segment a large number of low-resolution sample images obtained in step S11 into blocks to obtain image blocks, and the length and width of the image blocks are the same as step S21c;

S22b: Using the edge extraction algorithm to extract the edge information of the image block, counting the amount of information of each image block and the distribution of the amount of information of all image blocks;

S22c: Select the image block with the highest amount of information, obtain the pixel value of the image block as F1, take f=F1/4, then f*40%<=threshold<=f*60%, take several representative thresholds within this range For example, the amount can be taken as the following representative thresholds: f*40%, f*45%, f*50%, f*55%, f*60%, and calculate the sparse representation image super-resolution based on learning corresponding to each threshold point The reconstruction time of the high-rate reconstruction algorithm and the resolution of the reconstructed image can be judged according to the subjective evaluation and PSNR. Then determine the fixed threshold from several representative thresholds according to the user's needs. If the user needs to pay more attention to time, then increase the ratio, and the time will be faster; on the contrary, if the user pays attention to the reconstruction effect, then lower the ratio, which will bring Reconstruction works better but takes longer. For example, after calculation, f*50% best meets the needs of users, and then the fixed threshold=f*50%.

S22d: Input the low-resolution image block to be reconstructed obtained in step S21, and use the edge extraction algorithm to extract the edge information of the image block. When the information amount of the image block does not exceed the fixed threshold determined in step S22c, the image block is a low-information image block ; Otherwise, it is a high-information image block. Preferably, the edge extraction algorithm is a Canny operator edge detection algorithm.

S23: Perform image filtering operations on high-information image blocks, and extract high-frequency features, using a two-dimensional filter operator filter bank, the filter bank used is f={f ₁ , f ₂ , f ₃ , f ₄ }, It consists of four different filters, namely:

f ₁ =[1,-1], f ₂ =f ₁ ^T

(4)

f ₃ =LOG, f ₃ =f ₃ ^T

(5)

The superscript T represents the matrix transpose operation, and LOG represents a 5×5 two-dimensional filter operator. After the high-frequency feature extraction operation, the high-frequency information of the image block is obtained, represented by x _l .

S24: Perform KPCA dimensionality reduction on the image blocks obtained in step S23 to achieve high-dimensional data compression. The dimensionality reduction steps are as follows:

S24a: Express the high-dimensional data set as X={x ₁ ,x ₂ ,x ₃ ,…,x _M }, x _i ∈ R ^D , the KPCA method goes through a nonlinear mapping function x→Φ(x)∈F, where F is the feature space, so that each data x can be mapped to a high-dimensional feature space.

S24b: The kernel function will perform a corresponding operation from a point x to F through Φ, and the obtained F data meets the centralization condition, namely:

Then the covariance matrix in the feature space F is:

S24c: Find the eigenvalue λ≥0 and eigenvector of c

V∈F\{0}, Cv＝λv (8)

then there is

(Φ(x _v )·Cv)=λ(Φ(x _v )·v) (9)

Considering that all eigenvectors can be expressed as a linear combination of Φ(x ₁ ), Φ(x ₂ ),…,Φ(x _M ), namely:

Then there are:

In the formula, v=1,2,3,...,M, define M×M dimensional matrix K _μv

K _μv :＝(Φ(x _μ )·Φ(x _v )) (12)

S24d: Solve the above formula to get the eigenvalues and eigenvectors, and the projection of the data set in the eigenvector space V ^k can be written as:

Then, the data is projected onto the eigenvector V ^k of the covariance matrix, and the projection result (that is, the representation y of low-dimensional data) can be expressed as:

S25: Carry out learning-based sparse representation image super-resolution reconstruction on the data after step S24 dimensionality reduction and compression, the specific steps are as follows:

S25a: using the OMP algorithm for the sparse representation coefficient β of the low-dimensional data y obtained in step _S24d under the low-resolution dictionary D1, that is, solving the following equation:

where T ₀ is a given sparsity, and β _i is a sub-element in matrix β.

S25b: Multiply the obtained sparse representation coefficient β by the high-resolution dictionary D _h to obtain the reconstructed high-information reconstruction sub-region, namely:

X＝D _h β

Where X is the obtained super-resolution sub-region.

S26: Reconstruct the low-information image block obtained in step 22d using a bicubic interpolation algorithm to obtain a low-information reconstruction sub-region;

S27: Perform image stitching on the high-information quantum region and the low-information reconstruction sub-region to obtain a super-resolution reconstructed image;

S28: Perform M times interpolation on the super-resolution reconstructed image by using a bicubic interpolation algorithm to obtain an N*M (N=2, 3, 4; M>0) times super-resolution reconstructed image.

The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be assumed that the specific implementation of the present invention is limited to these descriptions. For those of ordinary skill in the technical field of the present invention, without departing from the concept of the present invention, some simple deduction or replacement can be made, which should be considered as belonging to the protection scope of the present invention.

Claims (10)

1. A reconstruction method of stepless scaling super-resolution image, is characterized in that, comprises the steps: Step 1, blurring and downsampling the high-resolution sample image according to the degradation model to obtain a low-resolution sample image; Step 2, using the K-SVD method to perform dictionary training using the low-resolution sample image to obtain a high-resolution dictionary and a low-resolution dictionary; Step 3, preprocessing the low-resolution image to be reconstructed to obtain several low-resolution image blocks; Step 4, using the low-resolution sample image to calculate a fixed threshold; Step 5, judging whether the pixel value of each of the low-resolution image blocks is smaller than the fixed threshold; if so, then judging that the low-resolution image block is a low-information image block; Low-resolution image blocks are high-information image blocks; Step 6: Reconstruct the high-information image block using a dictionary learning and sparse representation algorithm to obtain a high-information reconstruction sub-region; Step 7, reconstructing the low-information image block using an interpolation algorithm to obtain a low-information reconstruction sub-region; Step 8, performing image stitching on the high-information quantum region and the low-information reconstruction sub-region to obtain a super-resolution reconstruction image; Step 9: Perform M-fold interpolation on the super-resolution reconstructed image using an interpolation algorithm to obtain an N*M (N=2, 3, 4; M>0)-fold super-resolution reconstructed image. 2. the reconstruction method of stepless scaling super-resolution image according to claim 1, is characterized in that, described step 2 comprises the steps: (21) Extracting image features of the low-resolution sample image by a feature extraction method to obtain high-resolution feature information and low-resolution feature information of the space object. (22) Using the K-SVD method, perform joint training on the high-resolution feature information and the low-resolution feature information to obtain the high-resolution dictionary and the low-resolution dictionary. 3. the reconstruction method of stepless scaling super-resolution image according to claim 2, is characterized in that, described step (22) comprises the steps: (221) Utilize sparse K-SVD method to train low-resolution dictionary; (222) Compute a high-resolution dictionary. 4. the reconstruction method of stepless scaling super-resolution image according to claim 1, is characterized in that, described step 3 comprises the steps: (31) Denoising the low-resolution image to be reconstructed to obtain a first image; (32) Deblurring the first image to obtain a second image; (33) Segment and save the second image according to a fixed length and width to form the low-resolution image block. 5. the reconstruction method of stepless scaling super-resolution image according to claim 1, is characterized in that, described step 4 comprises the steps: (41) performing block segmentation on the low-resolution sample image to obtain a low-resolution sample image block; (42) Using an edge extraction algorithm to extract edge information of the low-resolution sample image blocks, counting the pixel values of each of the low-resolution sample image blocks and the distribution of pixel values of all the low-resolution sample image blocks; (43) Selecting X (X>1) candidate thresholds according to the pixel values of each of the low-resolution sample image blocks and the distribution of pixel values of all the low-resolution sample image blocks; (44) Utilizing a learning-based sparse representation image super-resolution reconstruction algorithm for each of the candidate thresholds to form a plurality of high-resolution candidate images; (45) Select the information amount threshold from X candidate thresholds according to the calculation time of the learning-based sparse representation image super-resolution reconstruction algorithm and the resolution of the correspondingly formed high-resolution candidate image. 6. the reconstruction method of stepless scaling super-resolution image according to claim 5, is characterized in that, described edge extraction algorithm is Canny operator edge detection algorithm. 7. the reconstruction method of stepless scaling super-resolution image according to claim 1, is characterized in that, described step 6 comprises the steps: (61) performing image filtering processing on the high-information image block, and performing high-frequency feature extraction to obtain high-frequency information; (62) Using the high-frequency information to perform KPCA dimensionality reduction on the high-information image block to realize high-dimensional data compression; (63) Using the OMP algorithm to reconstruct the data after dimensionality reduction and compression to obtain the reconstruction sub-region with high information content. 8. the reconstruction method of stepless scaling super-resolution image according to claim 7, is characterized in that, described image filtering operation adopts two-dimensional filtering operator filter bank, and described filter bank is f={f ₁ ,f ₂ ,f ₃ ,f ₄ }, where, f ₁ =[1,-1], f ₂ =f ₁ ^T f ₃ =LOG, f ₃ =f ₃ ^T T is a matrix transpose operation, and LOG is a 5×5 two-dimensional filter operator. 9. the reconstruction method of stepless scaling super-resolution image according to claim 7, is characterized in that, described step (63) comprises the steps: (631) Calculate the sparse representation coefficient β according to the OMP algorithm, wherein, Among them, y is the data after dimension reduction and decompression, T ₀ is a given sparsity, and β _i is a sub-element in matrix β; (632) Multiply the sparse representation coefficient β by the high-resolution dictionary to obtain the high-information reconstruction sub-region. 10. The reconstruction method of infinitely zoomed super-resolution images according to claim 1, wherein the interpolation algorithm is a bicubic interpolation algorithm.