Colorization by Patch-Based Local Low-Rank Matrix Completion

Authors: Quanming Yao, T. Kwok James

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on a number of benchmark images demonstrate that the proposed method outperforms existing approaches.
Researcher Affiliation Academia Department of Computer Science and Engineering Hong Kong University of Science and Technology Hong Kong {qyaoaa, jamesk}@cse.ust.hk
Pseudocode Yes The proposed procedure, which will be called Patchbased Local Low-Rank colorization (Pa LLR) in the sequel, is shown in Algorithm 2. ... Fast Inv(R) using Algorithm 3...
Open Source Code No The paper does not provide concrete access to source code for the methodology described. It only mentions that 'Codes for LCC and GLR are obtained from their authors', referring to baseline methods, not their own.
Open Datasets Yes Experiments are performed on eight color images from the Berkeley segmentation data set (Figure 3).
Dataset Splits No The paper states 'Varying numbers of pixels (1% 10%) are randomly sampled from the color image as observed labels input to the colorization algorithm', but does not provide specific training/validation/test splits for the images in the dataset.
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper mentions 'MATLAB notation' but does not provide specific software dependencies with version numbers.
Experiment Setup Yes We fix the patch size r to 16, and use k = 50 patches in each Pa LLR group. ...For ℓ2,... we fix the proportionality constant to 5, and µ = 0.16.