The　RIME　algorithm　is　a　novel　physical-based　meta-heuristic　algorithm　with　a　strong　ability　to　solve　global　optimization　problems　and　address　challenges　in　engineering　applications.　It　implements　exploration　and　exploitation　behaviors　by　constructing　a　rime-ice　growth　process.　However,　RIME　comes　with　a　couple　of　disadvantages:　a　limited　exploratory　capability,　slow　convergence,　and　inherent　asymmetry　between　exploration　and　exploitation.　An　improved　version　with　more　efficiency　and　adaptability　to　solve　these　issues　now　comes　in　the　form　of　Hybrid　Estimation　Rime-ice　Optimization,　in　short,　HERIME.　A　probabilistic　model-based　sampling　approach　of　the　estimated　distribution　algorithm　is　utilized　to　enhance　the　quality　of　the　RIME　population　and　boost　its　global　exploration　capability.　A　roulette-based　fitness　distance　balanced　selection　strategy　is　used　to　strengthen　the　hard-rime　phase　of　RIME　to　effectively　enhance　the　balance　between　the　exploitation　and　exploration　phases　of　the　optimization　process.　We　validate　HERIME　using　41　functions　from　the　IEEE　CEC2017　and　IEEE　CEC2022　test　suites　and　compare　its　optimization　accuracy,　convergence,　and　stability　with　four　classical　and　recent　metaheuristic　algorithms　as　well　as　five　advanced　algorithms　to　reveal　the　fact　that　the　proposed　algorithm　outperforms　all　of　them.　Statistical　research　using　the　Friedman　test　and　Wilcoxon　rank　sum　test　also　confirms　its　excellent　performance.　Moreover,　ablation　experiments　validate　the　effectiveness　of　each　strategy　individually.　Thus,　the　experimental　results　show　that　HERIME　has　better　search　efficiency　and　optimization　accuracy　and　is　effective　in　dealing　with　global　optimization　problems.