Publication figures

All five selected papers use original figures or a complete original subfigure extracted from the author-supplied PDFs. The manuscripts themselves are not included in the site. Only surrounding page furniture, body text, and separate captions are outside the crops. No scientific elements, labels, arrows, results, or colors have been redrawn, removed, or generated.

Each homepage thumbnail is a proportionally resized 1200-pixel-wide WebP. The larger PNG linked from the thumbnail preserves the original figure crop at high resolution. Homepage previews fit within their column width and a 180-pixel height limit, without stretching or cropping. This keeps a portrait figure from adding excessive height to its publication row. Clicking opens the complete PNG; clicking it again restores the page. Figures do not form a navigable gallery.

Coordinates below are PDF points, origin at the top-left: [left, top, right, bottom]. PDF page numbers are one-based.

Paper Figure / PDF page Crop Render DPI WebP dimensions
QS-QIM-RRW Fig. 4 / page 6 [75.96, 200.04, 273.96, 313.08] 600 1200 x 685
GenPTW Fig. 2 / page 4 [52, 52, 560, 252] 450 1200 x 472
DNLMark Fig. 1 / page 3 [77.94, 69.84, 536.04, 263.88] 400 1200 x 508
FrPCET-RRW Fig. 2 / page 5 [43.92, 64.98, 279, 358.02] 400 1200 x 1496
STDM-RRW Fig. 2(b) / page 6 [324, 132.96, 567.96, 321.96] 600 1200 x 930

Sources and figure selection

  • QS-QIM-RRW: author’s final submitted version of the T-IFS paper. Figure 4 shows decoding-region equalization and the resulting quantization strips. It is more compact than the full embedding/extraction flowcharts.
  • GenPTW: author’s camera-ready version of the AAAI paper. Figure 2 shows the attention and spatial fusion modules, plus the joint extraction and localization branches.
  • DNLMark: author’s copy of the ICASSP paper. Figure 1 retains the complete architecture, including the image-noise and latent-noise branches and the frozen/trainable legend.
  • FrPCET-RRW: author’s published copy of the T-MM paper. Figure 2 shows all four optimized-QIM cases, illustrating the smaller quantization error from choosing adjacent intervals.
  • STDM-RRW: author’s published copy of the T-IFS paper. Complete original panel 2(b) shows the proposed multi-level quantizer and all four embedding-strength levels. An author-hosted copy is also public.

The generic receiver/reconstruction diagrams from the earlier preview are no longer selected. Figure identifiers remain here for provenance, with no added figure captions below the homepage thumbnails.

The short contribution sentences in _data/selected_publications.yml paraphrase the corresponding abstracts, method descriptions, contribution lists and conclusions. They introduce no numerical or universal robustness claims.