Submitted manuscript, 8.55 MB, PDF document
Research output: Working paper
Research output: Working paper
}
TY - UNPB
T1 - Example-based learning for single-image super-resolution and JPEG artifact removal
AU - Kim, Kwang In
AU - Kwon, Younghee
PY - 2008/8/1
Y1 - 2008/8/1
N2 - This paper proposes a framework for single-image super-resolution and JPEG artifact removal. The underlying idea is to learn a map from input low-quality images (suitably preprocessed low-resolution or JPEG encoded images) to target high-quality images based on example pairs of input and output images. Toretain the complexity of the resulting learning problem at a moderate level, a patch-based approach is taken such that kernel ridge regression (KRR) scans the input image with a small window (patch) and produces a patchvalued output for each output pixel location. These constitute a set of candidate images each of which reflects different local information. An image output is then obtained as a convex combination of candidates for each pixel based on estimated confidences of candidates. To reduce the time complexity of training and testing forKRR, a sparse solution is found by combining the ideas of kernel matching pursuit and gradient descent. As aregularized solution, KRR leads to a better generalization than simply storing the examples as it has been donein existing example-based super-resolution algorithms and results in much less noisy images. However, this mayintroduce blurring and ringing artifacts around major edges as sharp changes are penalized severely. A prior modelof a generic image class which takes into account the discontinuity property of images is adopted to resolve thisproblem. Comparison with existing super-resolution and JPEG artifact removal methods shows the effectivenessof the proposed method. Furthermore, the proposed method is generic in that it has the potential to be applied tomany other image enhancement applications.
AB - This paper proposes a framework for single-image super-resolution and JPEG artifact removal. The underlying idea is to learn a map from input low-quality images (suitably preprocessed low-resolution or JPEG encoded images) to target high-quality images based on example pairs of input and output images. Toretain the complexity of the resulting learning problem at a moderate level, a patch-based approach is taken such that kernel ridge regression (KRR) scans the input image with a small window (patch) and produces a patchvalued output for each output pixel location. These constitute a set of candidate images each of which reflects different local information. An image output is then obtained as a convex combination of candidates for each pixel based on estimated confidences of candidates. To reduce the time complexity of training and testing forKRR, a sparse solution is found by combining the ideas of kernel matching pursuit and gradient descent. As aregularized solution, KRR leads to a better generalization than simply storing the examples as it has been donein existing example-based super-resolution algorithms and results in much less noisy images. However, this mayintroduce blurring and ringing artifacts around major edges as sharp changes are penalized severely. A prior modelof a generic image class which takes into account the discontinuity property of images is adopted to resolve thisproblem. Comparison with existing super-resolution and JPEG artifact removal methods shows the effectivenessof the proposed method. Furthermore, the proposed method is generic in that it has the potential to be applied tomany other image enhancement applications.
M3 - Working paper
T3 - Technical Report
SP - 1
EP - 28
BT - Example-based learning for single-image super-resolution and JPEG artifact removal
PB - Max Planck Institute for Biological Cybernetics
ER -