2D-DCT
An orthogonal transform that maps images from the spatial domain to the frequency domain. It is widely used in image compression standards such as JPEG and often serves as the measurement basis in single-pixel imaging.
Timeline (last 90 days)
LIPS-PC parameterizes the measurement operator as a residual correction to the 2D-DCT basis matrix, i.e., D = D_base + Δ, where Δ is trained end-to-end jointly with a compact classifier via a cross-entropy loss.
At M=8, the operator learned by LIPS-PC outperforms the best fixed transform (such as 2D-DCT) by 9-15 percentage points on all three datasets.
2D-DCT is the two-dimensional discrete cosine transform, widely used in image compression standards such as JPEG; using DCT basis as measurement patterns is a common choice in single-pixel imaging because natural image energy is highly concentrated in the DCT domain.
In the LIPS-PC parameterized policy, the DCT basis provides an information-rich initialization starting point, and the residual term Δ only needs to learn small task-relevant adjustments, thereby accelerating convergence and improving performance at low budgets; learning from a random matrix or zero matrix would significantly increase optimization difficulty.
All Facts (4)
LIPS-PC parameterizes the measurement operator as a residual correction to the 2D-DCT basis matrix, i.e., D = D_base + Δ, where Δ is trained end-to-end jointly with a compact classifier via a cross-entropy loss.
50%UnverifiedAt M=8, the operator learned by LIPS-PC outperforms the best fixed transform (such as 2D-DCT) by 9-15 percentage points on all three datasets.
50%Unverified2D-DCT is the two-dimensional discrete cosine transform, widely used in image compression standards such as JPEG; using DCT basis as measurement patterns is a common choice in single-pixel imaging because natural image energy is highly concentrated in the DCT domain.
50%UnverifiedIn the LIPS-PC parameterized policy, the DCT basis provides an information-rich initialization starting point, and the residual term Δ only needs to learn small task-relevant adjustments, thereby accelerating convergence and improving performance at low budgets; learning from a random matrix or zero matrix would significantly increase optimization difficulty.
50%