--- title: 'Boundary-to-Region (B2R): Core Concepts' url: https://www.emergentmind.com/topics/boundary-to-region-b2r type: topic --- # Boundary-to-Region (B2R): Core Concepts Searching arXiv for the cited B2R-related papers and nearby work to ground the article. Boundary-to-Region (B2R) denotes a class of formulations in which a boundary, or a boundary-like constraint, is treated as the primary organizer of region-level inference. In the cited literature, this idea appears in several technically distinct forms: detected object boundaries act as the source of contextual signals aggregated into interior pixels in semantic segmentation; learned semantic boundaries gate long-range propagation within segments; boundary evidence is converted into separated instance proposals; signed distance maps become region-level supervision weights; shadow, boundary, and non-shadow content are reordered before state-space scanning; cost-to-go is reinterpreted as a fixed safety boundary in offline safe reinforcement learning; and user-specified geographic boundaries are transformed into token sets and induced subgraphs for elastic region representation [2110.14587] [1909.00179] [2006.02695] [2108.01405] [2411.03260] [2509.25727] [2503.07991]. The term therefore names a recurrent operational pattern rather than a single algorithm. ## 1. Conceptual scope Across these works, B2R consistently replaces undifferentiated global processing with a boundary-conditioned restriction on where information originates, how it propagates, or which samples supervise the model. In dense prediction, the restriction is spatial; in restoration, it is a sequence permutation; in safe RL, it is a conditioning semantics; in urban representation, it is a geometry-to-subgraph map; and in mathematical or mechanical settings, it concerns how boundaries determine feasible regions or their projections. | Domain | Boundary signal | Region-level operation | |---|---|---| | Semantic segmentation | Detected object boundaries | Context aggregation into interior pixels | | Boundary-aware propagation | Learned semantic boundary probability | Gated feature diffusion within segments | | Nucleus segmentation | Instance boundary probability map | Connected-component proposals after boundary subtraction | | Region-wise supervision | Signed distance or boundary-derived maps | Per-pixel, per-class loss weighting | | Shadow removal | Shadow mask and boundary windows | Region-typed sequence reordering for SSM scanning | | Offline safe RL | Fixed safety budget boundary | Region-wide supervision over safe trajectories | | Urban representation | Prompted geometry | Token-set extraction and subgraph embedding | This diversity also clarifies that “boundary” is not always a geometric contour. In BCANet and BFP, it is a semantic edge field; in ShadowMamba, a window containing both mask values is a “boundary window”; in offline safe RL, the boundary is the deployment-time cost budget $\kappa$; and in BPURF, the boundary is a prompted polygon or related geometry [2110.14587] [1909.00179] [2411.03260] [2509.25727] [2503.07991]. ## 2. Boundary-guided dense prediction In "Boundary Guided Context Aggregation for Semantic Segmentation" [2110.14587], B2R is defined explicitly as the use of detected object boundaries as the source of contextual signals aggregated into interior pixels. BCANet consists of a Multi-Scale Boundary (MSB) extractor and a Boundary guided Context Aggregation (BCA) module. MSB taps the last residual block of each backbone stage in a dilated ResNet-101, unifies channels to $256$, resizes all scales to $1/8$ input resolution, and predicts a binary boundary map with a $1\times 1$ convolution and Sigmoid. BCA then performs a Non-local-style cross-stream attention in which semantic features $A$ and boundary features $B$ are projected to $A_1,B_1\in\mathbb{R}^{C\times N}$ and used to form a boundary-semantic affinity $$ F(i,j)=\frac{\exp(B_{1i}^\top A_{1j})}{\sum_{i'=1}^{N}\exp(B_{1i'}^\top A_{1j})}, $$ followed by the residual update $$ D_j = A_j + \sum_{i=1}^{N}F(i,j)\,A_{2i}. $$ The intended effect is that attention weights concentrate on boundary positions and push boundary-informed context into inner pixels, thereby increasing intra-class consistency and reducing inter-class confusion. On Cityscapes, relative to a dilated FCN baseline, BCANet improves interior F-score from $75.07$ to $77.02$ and boundary F-score from $57.93$ to $60.38$; BCANet with SegFix reaches boundary F-score $63.92$ and mIoU $80.86$. On the Cityscapes test set, BCANet with ResNet-101 reports $81.7$ mIoU, and on ADE20K validation it reports $45.62$ mIoU and $82.35$ pixel accuracy [2110.14587]. A related but architecturally distinct formulation appears in "Boundary-Aware Feature Propagation for Scene Segmentation" [1909.00179]. Here the network learns “boundary” as an additional semantic class, producing a boundary confidence map $b$ from an $N+1$-class softmax. B2R is implemented through a multiplicative propagation gate rather than an attention map. The boundary confidence is converted to a propagation confidence $$ p_t = 1 - \beta f(\alpha b_t - \gamma), $$ with $\alpha=20$, $\gamma=4$, and learnable $\beta$, and feature propagation along Unidirectional Acyclic Graphs is updated as $$ h_t = g(U * i_t + W * (h_{t-1}p_{t-1}) + \delta). $$ Messages inside a region are therefore passed strongly, while messages that would cross a boundary are attenuated. The paper replaces expensive DAG-style pixel scans with UAGs that reduce the number of sequential steps from $H\times W$ to $H+W$ and reports nearly identical mIoU to DAGs on PASCAL-Context under matched settings. Quantitatively, BFP reports $53.6$ mIoU on PASCAL-Context, $74.1$ mIoU on CamVid, and $81.4$ mIoU on Cityscapes test, with large gains over FCN baselines and further gains from boundary-aware gating [1909.00179]. Taken together, these two papers establish the main dense-prediction reading of B2R: boundaries are not merely auxiliary edge targets, but explicit regulators of where region semantics should be gathered from. ## 3. From boundaries to instances, losses, and variational contours In crowded nucleus segmentation, "Boundary-assisted Region Proposal Networks for Nucleus Segmentation" [2006.02695] realizes B2R by converting boundary evidence into separated instance regions before refinement. BRP-Net predicts both a semantic nucleus map $P_{\text{seg}}$ and an instance boundary map $P_{\text{bnd}}$ using a Task-aware Feature Encoding network with task-specific streams and residual Feature Fusion Modules. Proposal generation follows a contour-subtraction rule: threshold the two maps, set boundary pixels in the semantic map to zero, and compute connected components of the remaining foreground. This produces disjoint coarse proposals $\{R_k\}$, which are then refined by proposal-wise instance segmentation networks. The paper stresses that the original nomenclature is “Boundary-assisted Region Proposal,” but also states that this procedure is conceptually a B2R process. BRP-Net reports on Kumar an AJI of $64.22$ and F1 of $84.23$, and on CPM17 Dice1/Dice2/AJI of $87.7/79.5/73.1$, while emphasizing robustness to the dilation-radius hyper-parameter that affects contour-based splitting [2006.02695]. A more abstract conversion of boundary information into region supervision appears in "Region-wise Loss for Biomedical Image Segmentation" [2108.01405]. The Region-wise loss is $$ L_{\text{RW}}=\frac{1}{NK}\sum_{i=1}^N \hat y_i^\top z_i, $$ where $z_i$ is a per-pixel, per-class RW map derived from labels alone. Under this framework, Active Contour and Boundary loss can be reformulated by appropriate choices of $Z$. For boundary-derived supervision, naive signed-distance maps are shown to induce optimization instability because gradient signs depend on differences $z_{ik}-z_{il}$ and can produce multiple negative components in $\partial L/\partial \phi_i$. The paper introduces a rectification principle: for a pixel whose true class is $k$, all non-true channels should share the same value, and the true-class weight should satisfy $z_{ik}\le b_i$. Its rectified Region-wise map sets outside values to $1$ and inside values to normalized negative distances, yielding bounded, sign-correct gradients without auxiliary regularization or schedules. Experiments on ACDC17, BraTS18, and KiTS19 show state-of-the-art or comparable Dice and Hausdorff performance, and the convergence study reports that RRW eliminates the heavy tail of failed runs observed with unrectified boundary maps [2108.01405]. " A Region-based Randers Geodesic Approach for Image Segmentation" [1912.10122] takes yet another route. Instead of using boundaries only to define an eikonal speed, it injects region homogeneity into a Randers metric $$ \mathcal{R}(x,\xi)=\alpha(x,\xi)+\beta(x,\xi), $$ where the drift term $\beta(x,\xi)=\langle \omega(x),\xi\rangle$ is derived from the region term of an active-contour functional. This transforms the minimization of a region-and-boundary energy into the solution of a Randers eikonal PDE. The resulting interactive pipeline estimates closed contours by concatenating piecewise minimal geodesics inside a tube around the current boundary. The paper presents this as a way to prevent shortcutting through interiors when edge cues alone are weak or cluttered. Although the paper does not use the term B2R explicitly, its formulation is a boundary-to-region segmentation engine in the precise sense that regional appearance influences boundary geodesics throughout the tube [1912.10122]. ## 4. Sequence ordering and boundary-aware scanning In low-level restoration, "ShadowMamba: State-Space Model with Boundary-Region Selective Scan for Shadow Removal" [2411.03260] defines B2R as a mask-driven sequence reordering mechanism for a Mamba state-space model. Given a binary shadow mask $M$, the image is partitioned into non-overlapping windows of size $s\times s$ with $s=8$, and each window is typed by $$ f(W)= \begin{cases} 0,& P(W)=\{0\}\\ 1,& P(W)=\{0,1\}\\ 2,& P(W)=\{1\}. \end{cases} $$ This yields non-shadow, boundary, and shadow window sets, which are concatenated in the order non-shadow $\to$ boundary $\to$ shadow. The resulting permutation shortens sequence distances among same-type pixels, especially boundary pixels, before horizontal, vertical, reverse-horizontal, and reverse-vertical scans. ShadowMamba embeds this BRSSM branch inside a U-Net with a global GSSM branch, a CSSM channel branch, and an Efficient Feed-Forward Network. The central claim is not merely that masks help, but that the specific boundary-region ordering improves semantic continuity and local coherence along shadow boundaries, where brightness discontinuities are most abrupt. This is reflected in the ablation results on ISTD: GSSM alone gives RMAE $6.83$, adding BRSSM reduces it to $4.80$, adding CSSM gives $4.63$, and adding EFFN gives $4.23$. A scan-type ablation further reports RMAE $6.24$ for a mask scan without local scan, $6.13$ for a local scan without region separation, and $4.23$ for the boundary-region scan. The full model reports on SRD ALL RMAE $3.87$, PSNR $33.63$, SSIM $0.965$; on ISTD ALL RMAE $4.23$, PSNR $31.79$, SSIM $0.960$; and on ISTD+ RMAE $(2.8/5.8/2.3)$ for ALL/S/NS [2411.03260]. This version of B2R is notable because the “region” is not produced by spatial aggregation after attention or propagation. It is produced first by a permutation, and only then processed by a linear-time SSM. The paper therefore treats B2R as an ordering prior for long-sequence modeling rather than as a message-passing mask [2411.03260]. ## 5. Boundary-to-Region in offline safe reinforcement learning "Boundary-to-Region Supervision for Offline Safe Reinforcement Learning" [2509.25727] shifts B2R from geometry to decision-making. Its starting point is the asymmetry between return-to-go and cost-to-go in a constrained MDP: $$ \pi^*=\arg\max_\pi \mathbb{E}_{\tau\sim\pi}[R(\tau)] \quad \text{subject to}\quad \mathbb{E}_{\tau\sim\pi}[C(\tau)]\le \kappa. $$ The paper argues that RTG is a flexible target, whereas CTG is a rigid feasibility boundary. Standard DT-style sequence models condition symmetrically on RTG and CTG tokens, which the paper identifies as a source of brittle token selection and sparse, near-boundary supervision. B2R resolves this by filtering to safe trajectories, $$ D_{\text{safe}}=\{\tau\in D\mid C(\tau)\le \kappa\}, $$ and then realigning cost-to-go so that every safe trajectory is conditioned on the same deployment-time boundary token: $$ \hat C_t'=\hat C_t + (\kappa - C(\tau)), $$ which guarantees $\hat C_0'=\kappa$ while preserving the temporal cost-decay profile. Training then uses the usual autoregressive behavior-cloning form, but with realigned CTG tokens and RoPE. Conceptually, boundary-only supervision over trajectories near the budget is replaced by region-wide supervision over the entire safe region under a fixed boundary condition. The paper also states formal guarantees. Under Safe-Aligned Data and Prediction-Error Bound assumptions, it proves $$ \Pr[C^{\text{B2R}}(\tau)\le \kappa] \ge 1-\exp\!\left(-\frac{(\delta-\sigma H)^2}{2HC_{\max}^2}\right), $$ and $$ \mathbb{E}[C^{\text{B2R}}(\tau)]\le \kappa-(\delta-\sigma H). $$ It further proves reward dominance over boundary-only supervision when the dataset contains an optimal safe trajectory strictly inside the feasible region. Empirically, B2R is evaluated on $38$ DSRL tasks and is reported to satisfy safety in $35$ out of $38$ tasks while achieving the highest reward in $20$ tasks. The paper attributes remaining failures in CarCircle1/2 and AntCircle to dataset limitations, specifically the tight coupling of high reward and high cost under scarce safe data [2509.25727]. This formulation is a strong reminder that B2R need not involve images at all. Here the “boundary” is a rigid budget, and the “region” is the set of all safe trajectories consistent with that budget token. ## 6. Boundary-defined regions in urban representation, mechanics, and graph theory In "Boundary Prompting: Elastic Urban Region Representation via Graph-based Spatial Tokenization" [2503.07991], B2R is the transformation of a user-specified geometry into an elastic region embedding. A boundary $b$ induces a token set $$ S_b=\{t_i\in V\mid \mathrm{loc}(t_i)\in b\}, $$ which is extracted by querying an R-tree over spatial tokens and then expanding to adjacent virtual tokens through a spatial-virtual index. The induced subgraph is embedded by type-wise SUM aggregation followed by CONCAT, and then updated by multi-channel message passing over structure, position, and neighbor relations. The paper’s online extraction algorithm runs in $O(M_q\cdot D_v)$ time, where $M_q$ is the number of spatial tokens inside the queried boundary and $D_v$ is the average number of virtual tokens linked to each spatial token. On dynamic-region tasks in NYC and CHI, BPURF reports, for example, MAE/RMSE/$R^2$ of $0.376/0.504/0.745$ on NYC crime and $0.381/0.531/0.717$ on CHI crash, and it reports $10\times$ to $30\times$ subgraph-extraction speedups over naive extraction [2503.07991]. A different boundary-to-region notion appears in "Regions of possible motion in mechanical systems" [1401.6280]. There, the region of possible motion is the projection $R=\pi(J_k)$ of an integral manifold onto a lower-dimensional configuration space, and the generalized boundary is the visible contour of the projection. The paper states the criterion $$ \pi(x)\in \partial_g R \iff \operatorname{rank}V(x)